Scientific Publications

SummaryIn the gait experiment, 19 healthy college students aged 18–21 years (10 male and 9 female) completed six over-ground walks covering a distance of four meters at self-determined slow, medium, and fast speeds in a motion-capture laboratory. Hexoskin supplied tri-axial acceleration at 64 Hz; its other sensor data were not analyzed. The experiment compared the eyeglass-camera system and the body-worn accelerometer with 20-camera optical motion capture for classifying gait speed. Separately, the dissertation describes a video-based Fuzzy Inference System that uses video features and identified objects to classify activities.

Key FindingIn the gait experiment, the eyeglass-camera system significantly discriminated gait speed and outperformed the vest-based accelerometer, while motion capture provided superior gait-speed discrimination. The glasses-derived frontal-plane data distinguished slow, medium, and fast speeds in all reported pairwise comparisons (all p < .05); the accelerometer's vertical-plane data did not (p = .26). Separately, a video-based Fuzzy Inference System reported 74% overall activity-classification accuracy, but the dissertation cautions against direct comparison with systems that produce one definitive action label.

SummaryThis study developed a method for examining time-resolved biometrics in relation to climbing strategy, technique, and position on an indoor wall. Continuous Hexoskin measurements and smartphone video were time-matched so heart rate, breathing rate, minute ventilation, and hip acceleration could be viewed at specific positions along the route. An interactive interface supported review of biometric changes at key events and calculation of summary statistics. The illustrated demonstration was a 530 s (8:50, min:sec) lead climb on a 16.8 m (55 ft) indoor wall.

Key FindingDuring a 530-second lead climb monitored with Hexoskin, heart rate consistently increased during active climbing and decreased during rest, while breathing rate varied less consistently. Hip-acceleration peaks up to 0.36 g corresponded to active climbing, and sustained values of 0.01–0.08 g corresponded to rest. In a difficult 46–151 s section, heart rate, breathing rate, and minute ventilation all rose substantially; these descriptive results support the feasibility of linking sensor data with climbing video, and they do not show that a training intervention improved performance.

SummaryAbid and colleagues investigated inter-subject human-activity recognition from acceleration data collected with a Hexoskin shirt. They compared three classification pipelines based on handcrafted features with ReliefF and SVM, LDA with KNN, and a multivariate CNN, then combined them in an ensemble for ten static, dynamic and hazardous activities. The ensemble reported weighted-average precision, recall and F1 scores of about 99%, while parallel processing shortened segmentation and GPU execution accelerated the CNN under the reported setup.

Key FindingFor activity recognition from Hexoskin acceleration, the multivariate CNN and the three-pipeline ensemble each reported weighted-average precision, recall and F1 scores of about 99% across their evaluated activity classes. The ensemble covered ten activities, and under the reported setup, parallel processing made segmentation almost five times faster while GPU execution made the CNN about 80 times faster than CPU execution.

SummaryThis single-participant preprint evaluated a peak-space motion-artifact cancellation method for waist textile ECG during outdoor walking and jogging. A 54-year-old man completed four sessions on different days, each with approximately 20–25 minutes of walking and jogging. Hexoskin chest ECG served as the reference for waist ECG recorded with a Zephyr Bioharness.

Key FindingProcessed waist ECG generally tracked the Hexoskin chest-reference heart rate during walking and jogging, but tracking deteriorated during two approximately five-minute poor-quality periods. Artifact cancellation was less complete during walking than jogging. A secure waistband fit remained necessary because the method could not recover QRS complexes fully obscured by motion artifact; results come from one participant, and RMSSD agreement applies only to acceptable-quality segments.

SummaryAhmed and colleagues developed InterViewR to address limitations in digital interview training for individuals with Autism by combining an immersive virtual-reality interview with a Hexoskin smart shirt. Hexoskin captured RR intervals and breathing rate so the platform could present real-time affective feedback and help users recognize stress during different parts of an interview. The user evaluation examined feasibility, usability, and perceived usefulness, and the authors reported that the mixed-reality platform elicited genuine stress responses while providing coherent feedback on interview skills.

Key FindingInterViewR combined immersive virtual-reality interview practice with Hexoskin-based affective feedback, and the authors reported that it elicited stress responses and provided coherent feedback on interview performance. In a separate 11-person usability evaluation, none of the participants had Autism; they considered the system useful for interview training but did not expect to use it frequently. Suggested improvements included a more human-like interviewer, more varied settings, clearer feedback and objective performance scoring.

SummaryThis single-participant case report examined DFA a1, a nonlinear heart-rate-variability index, during cycling in a 62-year-old male recreational cyclist. It compared progressive five-minute intervals with and without 25 mg Atenolol and longer 10-minute intervals near the first ventilatory threshold. Hexoskin supplied heart-rate and R–R interval recordings, which were analyzed alongside ventilation and cycling workload.

Key FindingHeart rate measured with Hexoskin was 15–20 bpm lower at each power stage with Atenolol, while DFA a1 declined similarly with increasing cycling power in the control and Atenolol trials. At roughly 125 bpm, DFA a1 was 0.336 during Atenolol cycling and 0.723 during the control ramp at different workloads. These observations distinguish matching heart rate from matching workload in this individual; they do not establish universal training thresholds.

SummaryTen professional golfers completed 3-metre putting tasks under control, lavender, citrus, and ammonia conditions. A Hexoskin biometric shirt recorded heart-rate changes, while a force plate measured centre-of-pressure area. Olfactory inhalation produced no significant difference in putting success or heart rate, but postural-stability differences were significant in the address, downswing, and total putting sections.

Key FindingIn ten professional golfers, olfactory inhalation did not significantly change putting success (p = 0.115) or heart rate recorded by Hexoskin during the address, putting, or total phases (p = 0.838, 0.878, and 0.666). Postural stability measured by centre-of-pressure area was significantly inhibited in the address, downswing, and total putting sections (p = 0.000, 0.035, and 0.047).

SummaryThis protocol describes an exploratory single-subject study of yoga therapy after cancer treatment. Twenty-five adults were recruited and six dropouts were reported. The program comprised one individual yoga-therapy session and six weekly group sessions, with seven assessments from baseline through follow-up. Hexoskin shirts recorded ECG and respiratory data at assessment visits, and hierarchical linear modeling was planned for heart-rate variability and patient-reported outcomes. The publication reports a protocol rather than efficacy results.

Key FindingThis yoga therapy protocol used Hexoskin to monitor participants after cancer treatment; it reports no efficacy results. It describes a single-subject design with 25 recruited adults and six dropouts, using heart-rate variability as the primary outcome and fatigue, anxiety, depression, stress, perceived cognitive function and quality of life as secondary outcomes.

SummarySix adults with chronic obstructive pulmonary disease completed stair-stepping tests before and after bronchodilator treatment while wearing Hexoskin. Researchers compared the shirt with a Vmax metabolic cart; poor shirt signal quality excluded one participant from the main analysis. Tidal volume, minute ventilation and respiratory rate correlated between devices, but inspiratory capacity and inspiratory reserve volume did not. Changes after bronchodilator treatment did not correlate significantly between devices.

Key FindingIn five analyzable participants with COPD, Hexoskin tracked changes in breathing pattern during stair stepping, but the comparison with Vmax did not support using the shirt to detect bronchodilator-induced changes in operating lung volumes.

SummaryThis exploratory single-subject study analyzed 20 of 25 enrolled adults after cancer treatment. Participants received one individual yoga-therapy session and six weekly small-group sessions, with patient-reported fatigue, anxiety, cognitive function, depression, stress, and quality of life measured across phases. Hierarchical linear modeling compared changes over time. The article reports preliminary patient-reported outcomes rather than garment-derived physiological findings.

Key FindingThe authors reported that cancer-related fatigue and state anxiety increased immediately after the individual session but improved more during the subsequent group-intervention phase. Trait anxiety, perceived cognitive impairment, its impact on quality of life, and functional wellbeing also improved over time, while the remaining patient-reported outcomes did not change. These exploratory findings require confirmation in controlled trials and do not provide garment-derived physiological outcomes.

SummaryThis exploratory study combined psychological surveys with Hexoskin HRV monitoring in Australian early childhood educators. One hundred educators were recruited, and 68 had usable physiological data after cleaning. A common four-hour window from 10:00 a.m. to 2:00 p.m. was analyzed across work shifts; it was not a uniform total wear duration.

Key FindingIn a study pairing questionnaires with Hexoskin heart-rate variability monitoring, educators reported poorer functioning than normative data for emotional role limitations, energy/fatigue, and social functioning (BF10 > 100); this self-report comparison does not establish a physiological burnout marker. The mental-health association with RMSSD was small (τb = 0.17, p < .05) with anecdotal Bayesian support (BF10 = 2.25). Work–family conflict associations with RMSSD and HF power became nonsignificant after controlling for age, and the paper reported evidence against a relationship between burnout scores and physiological measures.

SummaryThis exploratory study followed four healthy young cyclists aged 25–37 years in Montreal over five days to examine traffic-related particulate-matter and noise exposure alongside pre- and post-cycling cardiac measures. Hexoskin supplied continuous ECG during 40-minute sitting periods and minute ventilation used with Aeroqual pollutant concentrations to estimate inhaled dose. Bayesian models analyzed 255 five-minute RR-interval segments.

Key FindingPM2.5 and PM10 doses estimated using environmental sensors and Hexoskin minute ventilation were highly correlated (r = 0.98), so the authors retained PM2.5 for subsequent analysis. Separate models reported associations of higher cumulative PM2.5 and noise dose with greater heart-rate variation, but the noise association disappeared when PM2.5 entered the combined model. The authors reported no HRV effect and cautioned against generalizing from four participants; the source's Bayes-factor interpretation is internally inconsistent, so no evidence-strength direction is assigned here.

SummaryMacQuarrie and colleagues compared Astroskin with the hospital-grade corpuls3 monitor in 24 participants during a seated 22-minute session; 19 returned for a second seated session. Heart rate and systolic blood pressure showed low error, strong concordance, and comparable between-session reliability. Oxygen-saturation error was low, but concordance and reliability were weaker, identifying SpO2 sensor placement and performance across a wider physiological range as areas for further study.

Key FindingAstroskin heart rate and systolic blood pressure showed low error, strong agreement with corpuls3, and comparable reliability across sessions. SpO2 error was low, but concordance and reliability were weaker, potentially reflecting limited variability and sensor placement.

SummaryFive orthopaedic surgeons wore Hexoskin garments during 179 joint-replacement operations to examine physiological workload across simple primary, conversion/complex primary, and revision knee and hip arthroplasty. The garment-derived outcomes were heart rate, minute ventilation, and energy expenditure. The patient case counts and demographics describe the operations, not the five monitored surgeons.

Key FindingIn five surgeons monitored with Hexoskin across 179 joint replacements, energy expenditure was descriptively higher during conversion/complex primary and revision arthroplasty than during simple primary procedures. After accounting for surgeon differences and patient BMI, complex primary knee arthroplasty remained higher than simple primary knee arthroplasty (β = 166.2, SE 60.7; P = 0.0073), whereas the corresponding hip comparison was not significant (P = 0.17). Heart rate and minute ventilation showed no significant overall group differences.

SummarySherman's system-safety study examined a crawling training device used by East Alabama fire departments, where reducing training-related injury risk was the central objective. The analysis proposed Hexoskin for remote physiological monitoring alongside infrared cameras and modeled it as a failsafe in the simulator fault tree. The model estimated that these additions would reduce the failure-mode detection score to 4 and the risk-priority number from 144 to 72; this was a modeled result, not a prospective validation with firefighter trainees.

Key FindingThe safety model estimated that adding Hexoskin and infrared cameras would reduce the simulator's failure-mode detection score to 4 and its risk-priority number from 144 to 72. This was a modeled safety-analysis result rather than a prospective validation of Hexoskin with firefighter trainees.

SummaryUsing Hexoskin thoracic and abdominal respiratory movement, this preliminary study developed five breathing gestures as inputs for a game-like Android application. Seven users tested the gestures and rated the experience an average 9 out of 10. The study identified the need for more personalized detection and calibration to accommodate individual breathing patterns.

Key FindingOf five interface gestures based on Hexoskin chest and abdominal movement, breath holding and thoracic sighing received the highest ease ratings. In a separate 'worked well' question, five of seven participants selected hold breath and five selected deep thoracic breath. Several gestures required more than one detection attempt, and participants reported difficulty distinguishing thoracic from abdominal breathing; the authors identified blurred gesture distinctions and repeated attempts as reasons for further refinement and personalized calibration; these were future work, not demonstrated improvements.

SummaryAlizadeh and colleagues used a 77 GHz frequency-modulated continuous-wave radar in a bedroom setting to estimate breathing and heart rates from a person lying on a bed. Hexoskin served as the wearable reference sensor for both measures. Radar estimates correlated 94% with Hexoskin breathing rate and 80% with Hexoskin heart rate under the reported conditions, so these results evaluate the radar rather than establish a new Hexoskin accuracy estimate.

Key FindingNon-contact radar estimates correlated 94% with Hexoskin breathing rate and 80% with Hexoskin heart rate. Hexoskin functioned as the wearable reference sensor, so the reported correlations evaluate the radar estimates rather than establish a new Hexoskin accuracy estimate.

SummaryThis study examined how intolerance of uncertainty and task appraisal relate to stress and performance in RCMP law-enforcement students. Participants completed questionnaires and simulated emergency-driving and use-of-force tasks. Hexoskin recorded heart rate as a measure of physiological arousal.

Key FindingAmong RCMP law-enforcement students, greater intolerance of uncertainty was associated with lower combined driving performance in the second session. Intolerance of uncertainty was also positively associated with task appraisal, while subjective stress was not significantly associated with the reported performance scores or maximum heart rate. These are correlational findings and do not show that either construct caused changes in stress or performance; the paper's abstract gives a broader null conclusion that conflicts with the detailed driving result, so the mixed findings should be interpreted cautiously.

SummaryAlizadeh and colleagues developed a contactless MIMO-radar system to monitor breathing and sleep position for two adjacent bedbound subjects at a time. The study reports three experimental rounds with two subjects per round, but it does not state that these were six distinct people. Hexoskin served as the breathing-rate reference. The radar reached 97% overall breathing-rate accuracy, a maximum RMSE of 1.7 breaths/min and more than 83% correct sleep-position detection under the reported conditions.

Key FindingUsing Hexoskin as the breathing-rate reference, the contactless radar achieved 97% overall breathing-rate accuracy and a maximum RMSE of 1.7 bpm across the reported measurements. Sleep position was classified correctly in more than 83% of cases, supporting simultaneous breath and position monitoring of two nearby sleepers under the tested conditions.

SummaryResearchers used Hexoskin vests and Actigraph watches to collect physiological and movement data for classifying self-reported classroom stress in five undergraduates. They compared several machine-learning approaches, including models that preserved the timing of observations. A temporal random-forest analysis with a 30-day sliding training window reported about 92% median accuracy and F1, but performance was weaker when person-specific models were tested on a later semester. The small, demographically narrow sample limits generalization.

Key FindingIn five undergraduates monitored with Hexoskin vests and Actigraph watches, a temporal random-forest model using a 30-day sliding training window reached about 92% median accuracy and F1. That result came from temporal cross-validation; person-specific models performed poorly when trained on one semester and evaluated on a later semester. The study did not establish performance in larger or more diverse groups; stress labels were self-reported, sensor coverage was limited, and factors such as sleep, caffeine and outside stress were not controlled.

SummaryThis study evaluated whether machine-learning models developed with wearable data from controlled conditions could extend to real-world anxiety detection in young adults. Hexoskin contributed ECG, respiration, and 3-axis acceleration data within a multimodal system that also used an E4 wristband. XGBoost and Decision Tree performed strongly when models were trained and tested on the same subsets, whereas Random Forest led transfer learning in calibration and in-lab settings and SVM led for in-the-wild data. The findings show that environmental noise, multimodal signal contributions, and the balance of feature importance affect model robustness and transferability.

Key FindingWith multimodal inputs including Hexoskin ECG, respiration and acceleration, XGBoost reached 0.95 accuracy and 0.94 F1 score on the combined calibration subset and 0.92 accuracy and 0.90 F1 score on WEAR in-lab data. In transfer learning, Random Forest performed best for calibration and in-lab data, while SVM led for in-the-wild data, so the baseline leaders did not consistently retain their advantage. The RADWear in-the-wild data were 80% not anxious, and the researchers undersampled that majority class while preserving the minority observations.

SummaryThis thesis combined a health survey, field monitoring, a fitness substudy and laboratory validation to examine paramedic health and work demands. In the field study, Hexoskin recorded heart rate, respiratory rate, physical activity, energy expenditure and steps during 232 shifts completed by 32 paramedics. A separate laboratory study compared Hexoskin heart-rate and respiratory-rate measurements with standard measurements during cycling. Participant counts and results belong to distinct study components and should not be combined.

Key FindingDuring 232 monitored shifts, the 32-paramedic field cohort spent more time on calls per shift in metropolitan postings than in regional postings. In a separate 14-paramedic fitness substudy, greater estimated aerobic fitness was associated with lower heart rate at shift start and across shifts. These observational results do not establish causation, and the fitness subgroup was self-selected; a cardiac-arrest analysis was limited to four paramedics and seven arrests.

SummaryThis thesis evaluated how human-activity-recognition models performed when wearable hardware or recording sessions changed, using Astroskin and Zephyr BioHarness accelerometer data collected from the author as the sole participant in a home setting after COVID-19 restrictions prevented the planned research. An Astroskin-trained model dropped from 99.07% accuracy on Astroskin data to 65.74% on BioHarness data, while another model declined from 99.57% to 50.95% when tested in a different Astroskin recording session. A separate out-of-domain discriminator identified data from unfamiliar hardware or recording sessions with 97.60% accuracy. The thesis proposes this discriminator as a warning when incoming sensor data no longer match a model’s training conditions, but the one-person home experiment does not establish broader performance.

Key FindingActivity-recognition accuracy declined when wearable hardware or recording sessions changed: Model A fell from 99.07% on Astroskin data to 65.74% on BioHarness data, while Model A Type 1 fell from 99.57% to 50.95% across Astroskin sessions. The BioHarness-trained Model H achieved 29.63% on Astroskin data, but the thesis gives conflicting values for its same-domain BioHarness result; when both hardware and session differed from training, accuracy reached 19.28%. The out-of-domain discriminator flagged unfamiliar device or session data with 97.60% accuracy, but the experiment used only the author as a participant in a home setting after COVID-19 restrictions prevented the planned research, so broader performance remains unestablished.

SummaryA randomized, double-blind crossover trial enrolled 23 girls aged 6–12 years with Rett syndrome. Participants received five days of oral ketamine or placebo, followed by a nine-day washout and the alternate treatment; the studied doses were 0.75 or 1.5 mg/kg per dose, given twice daily. Hexoskin contributed daily activity, sleep, heart and lung-function data as exploratory measures. The primary outcome was safety and tolerability.

Key FindingIn a crossover trial that used Hexoskin for exploratory monitoring, five days of low-dose oral ketamine did not improve clinical outcomes over placebo in girls with Rett syndrome, although exploratory EEG findings showed first-dose target engagement. Most related adverse events were mild or moderate; one participant stopped treatment after Grade 3 vomiting; only 10 participants contributed to the EEG analysis, and no correction was applied for multiple comparisons. The trial tested only the two lowest planned doses and stopped after two cohorts because of recruitment difficulties, limiting conclusions about higher doses or longer treatment.

SummaryAmelard and colleagues used Hexoskin smart-shirt cardiorespiratory signals and reference oxygen-uptake measurements from a metabolic system in 22 healthy young adults completing cycle-ergometer protocols. A temporal convolutional network modeled dynamic oxygen uptake across exercise intensities. The predicted oxygen uptake classified light, moderate and vigorous activity second by second with 94.1% accuracy across 16,090 seconds. The authors describe the approach as a path toward quantitative aerobic-activity monitoring outside the laboratory when combined with tidal-volume and heart-rate-reserve calibration.

Key FindingAcross cycling intensities, a temporal convolutional network using 218 seconds of Hexoskin cardiorespiratory data produced a combined oxygen-uptake error of −22 ± 122 ml min−1. The temporal models had lower limits of agreement and substantially fewer parameters than the stacked LSTM comparison, while a heart-rate-only configuration showed high bias and variance. Predicted oxygen uptake classified light, moderate and vigorous activity with 94.1% accuracy across 16,090 seconds.

SummaryNine male stock-car drivers were monitored during a competitive NASCAR event in hot conditions. Hexoskin continuously recorded heart rate during the racing session, while separate instruments measured core and skin temperature and environmental conditions. The study described hydration and physiological strain during a short race; it did not test whether these changes affected driving performance.

Key FindingDuring a roughly 17-minute race in which Hexoskin tracked heart rate and separate instruments measured temperature, nine male stock-car drivers lost 0.77 ± 0.3% of body weight on average and had a mean sweat rate of 0.63 ± 0.4 L/h. Core and skin temperatures rose, while the core-to-skin temperature gradient narrowed. The sweat-rate calculation included prerace activity, and the study did not measure the three- to four-hour fluid losses discussed as projections or test effects on driving performance.

SummaryCôté-Bélisle explored Hexoskin-supported biofeedback for emotional regulation and emotional awareness in young men with ADHD. Participants used heart-rate and breathing-rate reports to identify physiological-arousal surges and connect them with everyday emotional experiences. Both completers later reported greater competence in identifying, understanding, and managing their emotions.

Key FindingTwo young adults with ADHD completed the seven-week Hexoskin-supported biofeedback intervention. Both showed improved emotional regulation after the intervention and at the two-week follow-up, while emotional-awareness results varied by participant and instrument, with EDRE scores returning to baseline at follow-up. With two completers and no control group, the findings are preliminary and cannot establish that biofeedback caused the observed changes.

SummaryThis prototype compared two inverse-kinematics algorithms for augmented-reality digital twins of manual-handling tasks. Motion-capture joint positions served as the reference for avatar-joint error. Participants also wore a Hexoskin Shirt and Smart Device to supply heart rate, breathing rate and calorie data for the broader fatigue-analysis setup, but the study did not report a biometric or fatigue-detection validation.

Key FindingIn the tested digital-twin setup, Bio IK used 38 body points and produced lower avatar-joint position error than Unity IK using nine points. Bio IK required more processor time but slightly less memory, and the authors integrated it into an augmented-reality application. The comparison concerns avatar positions measured against motion capture; it does not validate Hexoskin biometrics or establish that one algorithm performs better for every movement or system.

SummaryThis paper reviews physiological monitoring aboard the International Space Station and describes Astroskin in the context of a student proposal for a space medical center. The proposed center would link crew monitoring with physician support.

Key FindingAstroskin is discussed as a potential source of physiological data. The paper reports a design proposal, with no new wearer measurements or clinical outcome.

SummaryThis engineering paper proposes a way to combine Bio-Monitor recordings with other astronaut health data and transmit them through space data relays. It describes how the service could support monitoring beyond low Earth orbit.

Key FindingThe proposed service would move beyond downloading data after wear. The paper describes a system design; it reports no test of real-time Bio-Monitor transmission or astronaut health outcomes.

SummaryThis dissertation examined associations between resting-state amygdala–ventromedial prefrontal cortex connectivity and emotional distress in daily life using functional MRI and 14 days of ecological momentary assessment. Thirty-two women completed the procedures; three left-handed participants were excluded from all analyses, leaving 29 right-handed women, and two participants with low random-prompt compliance were further excluded from momentary analyses, leaving 27. Right amygdala–ventromedial prefrontal cortex connectivity was associated with sadness instability and a behavioral-dysregulation composite, while overall mean negative affect and overall affective instability were not significantly associated with connectivity. Participants wore Hexoskin while awake to collect cardiac activity, respiration and accelerometry, but the dissertation explicitly excluded all physiological data from its analyses and reports no Hexoskin physiological results.

Key FindingAmong the 27 participants retained for momentary analyses, resting-state connectivity between the amygdala and ventromedial prefrontal cortex was associated with specific outcomes. Right-sided connectivity was associated with sadness instability and the behavioral-dysregulation composite, while binge eating was associated with connectivity on both sides; overall mean negative affect and overall affective instability showed no significant association. These are specific, non-causal fMRI and ecological-momentary-assessment associations: although participants wore Hexoskin during the 14-day awake data-collection period, the dissertation excluded its cardiac, respiratory and accelerometry data from analysis and reports no Hexoskin physiological results.

SummaryThis thesis developed data-driven models that estimate the full blood-pressure waveform from electrocardiogram and photoplethysmography signals and attach an uncertainty measure to each estimate. Astroskin supplied electrocardiogram and photoplethysmography signals for both study components and tri-axial acceleration for daily-activity monitoring. The daily-activity study used Portapres as the blood-pressure reference, while the cycling study used NOVA from Finapres Medical Systems; reference-device failures excluded six of 18 cycling participants, leaving 12 for analysis. The studies used distinct participant groups, so their results remain tied to their individual protocols.

Key FindingAcross five participants in a daily-activity analysis, complete blood-pressure waveform estimates had a mean absolute error of 6.18 ± 0.73 mmHg when walking data were included and 5.85 ± 0.73 mmHg when they were excluded. Separate exercise analyses found that nonlinear autoregressive models with external inputs (NARX) tracked large mean arterial pressure changes, but did not capture changes linked to peripheral resistance from ECG and forehead PPG alone.

SummaryResearchers used Astroskin ECG data to test a method for generating synthetic signals that combine features of wearable recordings with patterns from clinical ECGs. The study compared the generated signals with those from a model trained on mean squared error alone.

Key FindingThe authors reported that generated signals were 3.6 times closer to the original asymptomatic wearable ECGs and 1.3 times closer to the symptomatic clinical ECGs than the baseline model’s outputs. This measured signal similarity; it did not test diagnosis in astronauts.

SummaryThis study developed a mobile human-activity-recognition application using three-axis acceleration from a Hexoskin garment. The classifier was evaluated on data from 40 healthy volunteers who repeated 10 activities and achieved 95.4% overall accuracy with a support vector machine under leave-one-subject-out testing. Home cardiac rehabilitation was the intended application, but no rehabilitation patients or clinical outcomes were evaluated.

Key FindingThe support-vector-machine classifier reached 95.4% overall activity-classification accuracy under leave-one-subject-out testing. It used Hexoskin acceleration data from 40 healthy volunteers, not cardiac-rehabilitation patients. The application displayed recognized activities and included an archive of medical reports for doctors, but increased home-rehabilitation participation remained an aim rather than a demonstrated outcome.

SummarySaini evaluated machine-learning and deep-learning models for human-activity recognition using participant-labeled Hexoskin data. Models trained with Capture24 were tested on Hexoskin data for activities including walking and sleeping. The conclusion reports 98.4% accuracy for an LSTM configuration that did not require manual feature engineering to distinguish walking from sleeping. The thesis presents healthcare use as a future possibility rather than a demonstrated clinical benefit.

Key FindingThe Hexoskin evaluation set focused on sleep and walking annotations, and the classifiers were assessed with accuracy, recall, precision and F1 score. The LSTM without manual feature engineering produced the strongest balanced test result reported in the table: 98.4% accuracy, 97.5% recall, 93.4% precision and 95.4% F1 score. Several tree and boosting configurations showed train-test gaps consistent with overfitting, and the thesis does not establish effectiveness in healthcare settings.

SummaryThis thesis contains two separate human studies. A 42-minute cooling-garment test involved 19 male completers and evaluated a liquid-carbon-dioxide system in a hot, humid chamber. A second study compared Hexoskin and Polar H7 heart-rate measurements in 12 male volunteers during a 28-minute cycling protocol under moderate and hot, humid conditions. Cooling outcomes came from separate Polar H10 and thermometer measurements.

Key FindingIn the cooling study, which used Polar H10 for heart rate, wearing the garment was associated with lower internal body temperature and heart rate, lower perceived exertion and better comfort, humidity sensation and well-being during the 42-minute test. In the separate validation, average Hexoskin–Polar heart-rate correlations were 0.9987 in moderate conditions and 0.9985 in hot, humid conditions. These correlations do not establish individual interchangeability, and no miner field deployment was tested.

SummaryApparicio and colleagues monitored three healthy male cyclists over five dry days in Delhi while modeling road-level noise and nitrogen-dioxide exposure. Hexoskin measured heart rate, breathing rate and minute ventilation; minute ventilation was combined with Aeroqual nitrogen-dioxide concentration to estimate inhaled dose. For one hour of cycling under moderate predicted pollution, estimated noise dose was 212% on residential streets and 459% on primary roads, while estimated inhaled nitrogen dioxide was 403 µg and 482 µg, respectively.

Key FindingIn Delhi, researchers combined Hexoskin ventilation data with environmental measurements. Cycling on a primary road rather than a residential street was associated with 47 µg/m³ more NO2 and 3.3 dB(A) more noise. After one hour in areas with moderate predicted pollution, estimated noise dose reached 212% on residential streets and 459% on primary roads, while inhaled NO2 reached 403 µg and 482 µg, respectively.

SummaryA single-site pilot compared physiological stress in 16 pediatric emergency physicians during real emergency-department resuscitations and a one-hour virtual-reality session. Across 69 monitored shifts and 31 resuscitations, heart-rate and salivary-cortisol increases were larger in real events, while perceived workload was higher in virtual reality. Hexoskin measured heart rate; cortisol and workload were separate measures.

Key FindingHeart rate measured with Hexoskin and cortisol measured in saliva rose more during real emergency-department resuscitations than during virtual simulations. Virtual-reality scenarios still increased heart rate, but the study did not establish equivalent physiological stress. This was a small, single-site provider study: nine shifts had entirely missing heart-rate data, and resuscitation type and patient outcomes were not collected; the real and virtual phases occurred in different settings and periods, so the comparison does not isolate simulation format alone.

SummaryPelgröm used Hexoskin thoracic and abdominal respiratory-inductance signals to compare breathing coordination during maximal cycling in 12 people with asthma and 15 healthy controls. The study introduced BPratio, describing the balance between tidal volume and breathing frequency, and Ldis, a measure of thoraco-abdominal asynchrony. Most participants increased breathing depth before breathing frequency, while 33% of the asthma group accelerated breathing earlier in the workload. Deviant Ldis patterns were more common in asthma and were associated with reduced performance or greater post-exercise dyspnoea.

Key FindingMost participants increased tidal volume before accelerating breathing frequency late in exercise, but one third of the asthma group increased breathing frequency earlier, at about 30% to 40% of maximum workload. Both markedly increasing and unusually stable thoraco-abdominal asynchrony were associated with signs of dysfunctional breathing. These deviant patterns occurred in 33% of participants with asthma and 7% of healthy controls.

SummaryThis planned secondary analysis followed 21 pediatric emergency-medicine providers over 91 eight-hour shifts. Hexoskin recorded heart rate under work clothes, while salivary cortisol was sampled separately at scheduled points. Shifts with a preceding event had higher heart rate and cortisol at midshift and end-shift than shifts without an event.

Key FindingHeart rate measured with Hexoskin was higher at midshift and end-shift when a preceding event had occurred. Cortisol also remained higher at those points, rather than showing the decline seen on shifts without an event. Heart rate and cortisol were summarized at defined time points; the observational analysis did not establish how long changes lasted after shifts, cumulative health effects or patient outcomes.

SummaryThirteen healthy volunteers each repeated 10 physical activities five times while wearing a Hexoskin garment with a waist-level accelerometer. The study classified these activities plus an unknown transition class using k-nearest-neighbor and support-vector-machine models. Overall test accuracy was 96.37% for the k-nearest-neighbor model and 95.4% for the support-vector-machine model under a random sample-level split.

Key FindingUsing waist acceleration from Hexoskin shirts worn by 13 healthy volunteers, a k-nearest-neighbor model classified activities with 96.37% accuracy. The researchers split samples at random, so the result does not show how well the model would classify the movements of a new participant.

SummaryThirty-two adults with COPD or interstitial lung disease completed randomized four-minute treadmill tests with fan-to-face, fan-to-leg and no-fan conditions. Two participants could not wear the Hexoskin shirt and three more had unusable software data, leaving usable Hexoskin physiological data for 27 participants. Fan-to-face reduced breathlessness by 0.5 Borg units versus no fan but did not outperform fan-to-leg; ventilation, breathing pattern, heart rate and blood oxygen saturation did not differ across conditions. Clinically meaningful reductions of at least one Borg unit occurred in 31% of the full sample and 45% of the more-breathless exploratory subgroup under fan-to-face versus no fan.

Key FindingAcross all 32 participants, fan-to-face reduced breathlessness by 0.5 Borg units versus no fan but not versus fan-to-leg; the exploratory subgroup with at least moderate breathlessness improved by 0.9 units versus no fan. Usable Hexoskin physiological data were available for 27 participants, and ventilation, breathing pattern, heart rate and blood oxygen saturation did not differ across conditions.

SummaryThe study compared Hexoskin heart rate, breathing rate and estimated ventilation with simultaneous laboratory references during a progressive maximal-exercise test. Results were reported for 28 healthy participants. Heart-rate and breathing-rate measurements generally showed high agreement, while estimated-ventilation results varied too widely for a firm validity conclusion.

Key FindingReported agreement coefficients for Hexoskin heart rate versus the laboratory reference were 0.98–0.99 for almost all participants, but one participant had much lower values of 0.58 and 0.59 after sensor-contact problems. Breathing-rate coefficients ranged from 0.91 to 0.99, whereas estimated-ventilation coefficients ranged from 0.2 to 0.99 and remained difficult to validate. The healthy laboratory sample and progressive exercise protocol do not establish performance in clinical or daily-living settings.

SummaryIn a controlled dust chamber, the study monitored particulate matter and inhalation rate across 15 construction activities and materials. Alphasense OPC-N3 sensors measured PM1, PM2.5 and PM10, while a Hexoskin wearable supplied inhalation-rate measurements. The authors used these measurements in a hazard-index model; the results do not represent disease incidence or demonstrate risk reduction at real construction sites.

Key FindingThe exposure model, which combined particle measurements with Hexoskin inhalation rate, calculated the highest particle-size hazard levels for M20 and M25 concrete mixing, even though the highest particulate-matter emissions occurred during concrete-block cutting. Mixing also had higher inhalation rates than the other activities. These are modeled hazard indices from controlled dust-chamber tests, not observed disease rates or evidence that a control reduced exposure; field validation at construction sites remains needed.

SummaryIn controlled dust-chamber simulations, this study combined real-time inhalation-rate data with particulate-matter and toxic-substance measurements to model construction-activity health risks. Hexoskin measured inhalation rate, Alphasense OPC-N3 sensors measured particulate matter, and MiniVol sampling supported toxic-substance analysis. The resulting indices are modeled hazard estimates, not observed disease rates or evidence of reduced exposure.

Key FindingIn a drilling study that paired particle sensors with Hexoskin inhalation-rate measurements, particulate-matter concentration was highest for M25 concrete blocks at 4.151 mg/m³ and lowest for solid blocks at 1.685 mg/m³. The paper reported aggregate toxic-substance hazard indices above its stated threshold across activities, while particulate-matter indices remained below it. These modeled classifications do not establish disease risk, regulatory compliance or performance at construction sites.

SummaryResearchers tested whether recognizing a person’s activity from an ECG before verifying identity improved biometric performance. They analyzed two public datasets, including ECG recorded with a Hexoskin Hx1 in the Vollmer dataset. Classifying activity first generally improved verification compared with using the ECG directly; a Naïve Bayes verifier generally produced lower equal-error rates than a decision tree.

Key FindingIn a comparison that included Hexoskin ECG, identifying activity before checking identity improved biometric verification relative to direct verification. The result concerns the analysis method, not a clinical diagnosis.

SummaryThis study used a Hexoskin-derived dataset containing heart rate, breathing rate, minute ventilation, cadence and activity to build forecasting and action-state models. Random-forest and XGBoost models predicted heart and breathing rates up to 25 minutes ahead. A separate hardware prototype converted sustained predicted drops or spikes into haptic and sound alerts; the study did not demonstrate clinical or safety outcomes.

Key FindingThe stacked model for forecasting breathing rate from a Hexoskin dataset reported 8.1% on the study's SI error metric. One example trained on one minute of data and predicted the next four minutes. Support-vector-machine classification performed less well than other tested approaches; these results come from the study dataset and prototype and do not establish clinical accuracy, real-world hazard prevention or universal forecasting performance.

SummaryThis study investigated activity-aware electrocardiogram biometric verification using deep learning on wearable devices. It used ECG-only models in a framework designed to work with different device types, including a Hexoskin smart shirt with textile sensors. Activity-aware verification generally outperformed direct verification, although performance varied widely across image representations, window lengths and model settings.

Key FindingScalogram images generally produced the lowest equal-error rates, while Mel-spectrograms produced the highest; Faros performed better with SGDM and Hexoskin with Adam. The best reported Hexoskin result was a 0.49% equal-error rate in the four-second scalogram standing condition, with 0.49% false acceptance and 0% false rejection across one, three and five genuine enrollment samples. Activity-aware verification generally outperformed direct verification, but performance varied widely across model settings.

SummaryResearchers built a digital twin of an industrial assembly task using physiological measurements from a Hexoskin vest, camera images and movement data from instrumented gloves. The prototype used these inputs to model movement and fatigue signals and trigger feedback. The authors observed unrealistic changes in simulated biometrics and avatar movement, limiting what the simulation can establish about workers.

Key FindingIn a prototype simulation combining Hexoskin measurements with cameras and smart gloves, the system reported 85.4% accuracy for fatigue-signal detection. Reported errors for the capacitive heart-rate and breathing-rate estimates were 3.14 and 2.91, respectively; these were not accuracy measurements of the shirt. Unrealistic biometric changes and avatar motion limit the interpretation of the simulation.

SummaryThis preliminary study compared infant heart-rate and heart-rate-variability measures derived from electrocardiogram and RR-interval files after 5–10-minute resting recordings at one month of age. Researchers used a Hexoskin Shirt Junior, custom conversion scripts and Kubios software. Results differed for one variability measure, and only heart rate and low-frequency power were significantly correlated between file types.

Key FindingRoot mean square of successive differences showed a significant fixed bias between Hexoskin electrocardiogram-derived and RR-interval files, although Bland–Altman plots assessed several heart-rate-variability measures. Only heart rate and low-frequency power were significantly correlated between procedures. A discussion passage contradicts the latter result for low-frequency power, so these preliminary findings require cautious interpretation and standardized processing.

SummaryBaig and colleagues combined Hexoskin cardiac, respiratory, and activity measurements with blood glucose, demographics, clinical rules, and a knowledge base in an adaptive neuro-fuzzy model for early detection of prediabetes and type 2 diabetes. The approach was evaluated longitudinally in two people using 7.25 hours of real-time data and reported 91% accuracy, 94% sensitivity, 90% specificity, and 91% agreement with clinical assessment. The results support further evaluation of individualized wearable monitoring, while the two-person test limits generalizability.

Key FindingUsing Hexoskin cardiac, respiratory, and activity inputs, the adaptive neuro-fuzzy model achieved 91% accuracy, 94% sensitivity, and 90% specificity for the reported prediabetes and type 2 diabetes detection task; Kappa testing reported 91% overall agreement with clinical assessment.

SummaryA single surgeon performed 26 direct-anterior total hip replacements with manual femoral broaching and 20 with a battery-powered automated system. Hexoskin-derived surgeon energy expenditure, heart rate and minute ventilation were compared during broaching. The automated method was associated with shorter broaching time and lower measured physiological demand in this observational, single-surgeon comparison.

Key FindingFor one surgeon performing total hip replacements, automated broaching took 3.7 minutes on average versus 6.1 minutes with manual broaching. Hexoskin estimates of the surgeon’s energy expenditure averaged 16.0 versus 32.6 Calories, respectively.

SummaryBaig and colleagues combined Hexoskin vital-sign and activity measurements with blood glucose, demographics, clinical guidelines, and a knowledge base in an adaptive neuro-fuzzy model for early detection of prediabetes and type 2 diabetes. The model was trained with data from 36 participants and tested longitudinally in two people. It reported 91% overall Kappa agreement and included a descriptive two-year follow-up.

Key FindingA model combining Hexoskin measurements with clinical and other inputs achieved 91% overall agreement with expert assessment for early prediabetes and type 2 diabetes detection. In the two-person follow-up, the participant using the wearable and mobile-health approach had average HbA1c of 64 mmol/mol versus 87.5 mmol/mol in the participant receiving traditional management, but this descriptive comparison does not establish causality.

SummaryBanerjee and colleagues evaluated Hexoskin as a prospective unobtrusive monitoring platform for dementia care, first validating cadence in four healthy adults across rest, walking, running, and sprinting. Cadence changed in the expected order across the four activity states. This preliminary validation did not test people with dementia.

Key FindingHexoskin cadence increased in the expected order from rest to walking, running, and sprinting, with 95.5% of cadence variation explained by activity state. Breathing rate, minute ventilation, and accelerometer activity were significantly correlated with cadence, supporting their use for activity-pattern detection, whereas heart rate was the least consistent cadence correlate.

SummaryThis study developed a logistic-regression model to estimate manufacturing-operator engagement from respiration and acceleration data collected with Hexoskin. Under leave-one-out cross-validation, the model reported 80.95% accuracy. The prototype displayed engagement on a continuous color gradient calibrated to the range observed in its training data.

Key FindingOperator engagement was higher in manual work than in the automated condition on the Utrecht Work Engagement Scale absorption subscore, 4.21 versus 3.32. The logistic-regression model using Hexoskin respiration and acceleration predicted engagement states with 80.95% accuracy under leave-one-out cross-validation. The feedback display is a prototype based on respiration and acceleration; the study does not establish improved engagement, productivity or safety.

SummaryBanerjee and colleagues evaluated Hexoskin against laboratory reference measures in 31 adults aged 65 years or older, including participants with mobility impairment, cognitive impairment, both, or neither. Processed cadence, respiration rate, and heart rate showed useful agreement with their references, while minute ventilation and short-duration activity level did not. The findings support selective longitudinal monitoring in older adults and emphasize signal-specific validation before dementia-management use.

Key FindingIn 31 older adults, processed Hexoskin cadence, respiration rate and heart rate tracked their corresponding reference measures, while minute ventilation and short-duration activity level did not. Heart rate and processed cadence showed the strongest agreement, and respiration rate showed a moderate but significant association with low bias. The results support selective use of the validated signals and do not establish that every commercial-sensor output is clinically interchangeable with its reference.

SummaryTwenty-two participants completed manual and fully automated snowshoe-assembly tasks while wearing Hexoskin for physiological data collection. Those data were used to build the Optimal State Deviation Index for estimating engagement. The model reported 80.95% test accuracy under leave-one-out cross-validation. A separate prototype displayed engagement through a color gradient as part of an iterative biocybernetic-feedback design.

Key FindingThe Optimal State Deviation Index, developed using Hexoskin data during manual and automated snowshoe assembly, achieved 81.31% accuracy on training data and 80.95% under leave-one-out testing. The formula used the full Task 1 and Task 2 dataset, so the test estimate comes from cross-validation rather than a separate external cohort. In display testing, participants found the discrete color gradient more distracting; the study did not demonstrate improved workplace performance or safety.

SummaryFive high-volume surgeons were monitored with Hexoskin during 144 consecutive total hip arthroplasties: 67 direct-anterior and 77 posterior procedures. Direct-anterior cases were shorter but were associated with higher surgeon stress index, heart rate and minute ventilation. This observational comparison concerns surgeon physiology and does not establish patient outcomes or that surgical approach caused the differences.

Key FindingIn five surgeons monitored with Hexoskin, direct-anterior procedures had a mean stress index of 17.4 versus 12.4 for posterior procedures. Heart rate averaged 101 versus 98.3 beats/min, and minute ventilation 21.7 versus 18.7 L/min. Operative time was shorter for the direct-anterior approach; results came from five high-volume surgeons and may not generalize to other surgeons, settings or patient outcomes.

SummaryBarker and colleagues used Hexoskin's hip-mounted tri-axial accelerometer to compare seven male acrobats' trampoline performances across training and shows. Training produced higher average, RMS, and RM4 acceleration, whereas perceived exertion was higher during shows and time in defined acceleration ranges did not differ significantly. The study positions wearable inertial measurement as a practical workload-management tool for theatre and arena performances where GPS is inoperable.

Key FindingHexoskin accelerometry showed higher average, RMS, and RM4 acceleration during training than shows, while artists rated shows as more exerting and time within the five acceleration ranges did not differ significantly. These track-level profiles suggest wearable inertial data can help coaches characterize mechanical demands in theatre environments where GPS is unavailable.

SummaryNine healthy young men completed four randomly ordered exercise sessions in hot, dry conditions. Each session included seated rest, 60 minutes of semi-recumbent cycling at a fixed metabolic heat production and 60 minutes of recovery. Comparisons found no evidence that Hexoskin, worn alone or under a work uniform, impaired whole-body heat dissipation under the tested conditions.

Key FindingBody heat storage was similar with Hexoskin alone and the semi-nude control, 214 versus 211 kJ. Under a work uniform, storage was also similar with Hexoskin and a cotton undershirt, 307 versus 318 kJ. Evaporative, dry and total heat loss did not differ significantly between the paired conditions; this controlled study in nine men does not establish results for other populations, clothing systems or environments.

SummaryThis master’s thesis compared Hexoskin with a Vmax metabolic cart during incremental and constant-workload cycling in adults with chronic obstructive pulmonary disease. Fourteen participants wore the shirt; seven contributed to the assessment of change after pulmonary rehabilitation. Heart rate and breathing rate changed with exercise, without significant device or device-by-time effects across the three tests. The findings for tidal volume and inspiratory capacity were less consistent.

Key FindingHexoskin and Vmax showed similar heart-rate and breathing-rate responses during exercise in the COPD cohort. The small rehabilitation subgroup and the less consistent lung-volume results limit conclusions about using the devices interchangeably.

SummaryTwelve Division I collegiate athletes completed graded treadmill tests while wearing an N95 respirator, a cloth mask or no face covering. Hexoskin continuously measured heart rate and breathing rate. Exercise lasted longer without a face covering; heart-rate slopes were similar across conditions, while breathing rate increased more steeply without a covering than with a cloth mask.

Key FindingHeart rate measured with Hexoskin increased similarly across the N95, cloth-mask and no-covering conditions, with no significant difference in slopes. Breathing rate rose in all conditions, but the increase was steeper without a covering than with a cloth mask. Transcutaneous carbon dioxide decreased over time only in the no-covering condition; results from 12 athletes and a maximal treadmill protocol do not establish effects in other populations or everyday activity.

SummaryBeltrame and colleagues monitored 13 healthy active men with Hexoskin for about six hours per day over four days to study aerobic-system dynamics during unsupervised daily activity. Hexoskin cardiac, respiratory and movement features were used to predict free-living oxygen uptake and derive aerobic-system dynamics. The predicted daily-life temporal measure correlated with the directly measured laboratory measure at r=0.82 (p<0.001).

Key FindingAerobic-system mean normalized gain derived from predicted oxygen uptake during unsupervised daily activity correlated with the laboratory-derived measure at r=0.82 (p<0.001; n=13), with a consistent difference of about 16%. The model distinguished light, moderate and vigorous activity, but only about 20% of monitored data occurred while participants were active enough for reliable aerobic-system analysis. Hexoskin supplied continuous activity, breathing and heart-rate inputs, while free-living oxygen uptake was model-predicted and the proposed health-status application remained speculative.

SummaryA field study monitored 32 paramedics with Hexoskin around work shifts and emergency calls, recording heart rate, respiratory rate and occupational acceleration. Fourteen also completed maximal cardiorespiratory fitness testing. The study examined how posting, body mass, call characteristics and fitness were associated with physiological responses during paramedic work.

Key FindingIn 32 paramedics monitored with Hexoskin, posting, body mass index and call type were associated with physiological responses during shifts and emergency calls. Greater cardiorespiratory fitness was associated with lower physiological demand among the 14 participants who completed fitness testing. These observational findings do not show that posting, body mass or fitness caused the measured differences.

SummaryBeltrame and colleagues evaluated oxygen-uptake prediction in 16 healthy active men during controlled random-paced walking and approximately 20 minutes of simulated daily activities. Hexoskin cardiac, respiratory, and motion features supplied the wearable inputs, while a COSMED K4b2 system provided the directly measured oxygen-uptake reference. The findings support prediction under the tested activity conditions and identify the need for further reliability testing before broader application.

Key FindingIn 16 healthy active men, the Hexoskin-feature random-forest model predicted oxygen uptake during simulated daily activities with r=0.87 versus COSMED measurements and produced mean normalized gain correlated at r=0.71 with the measured endpoint. The model distinguished resting, light, and moderate metabolic demands, but intense activity represented less than 2% of the data and was excluded from the validated range.

SummarySixteen participants completed 20-minute standardized data-entry tasks in an active chair and a conventional chair. A metabolic cart measured oxygen uptake and ventilation, while Hexoskin measured heart rate and heart-rate variability. On average, active sitting increased oxygen uptake by 0.02 L/min and energy expenditure by 1.5 kcal per 15 minutes, with substantial individual variation.

Key FindingTen of 16 participants increased oxygen uptake in the active chair, two showed no change and four showed a decrease; the average change was 0.02 ± 0.04 L/min. Hexoskin cardiac measurements showed larger relative heart-rate-variability decreases in males than females, although absolute between-sex differences were not significant. This short laboratory study does not establish long-term health, productivity or weight-management benefits.

SummaryThis study used Hexoskin as a reference sensor to evaluate non-contact breathing-rate estimation with a frequency-modulated continuous-wave radar. A nonlinear least-squares estimator was compared with discrete Fourier transform and frequency-time phase-regression approaches in static, texting, and typing conditions. The proposed method produced high within-one-breath-per-minute accuracy in the laboratory tests.

Key FindingIn a 10-minute seated test, 94.5% of breathing-rate estimates from a 77 GHz radar and nonlinear least-squares model were within one breath per minute of the Hexoskin reference. Across the tested scenarios, the authors reported accurate estimation and improvement over discrete Fourier transform and frequency-time phase-regression methods, but the laboratory evidence does not establish performance in a broader population or free-living setting.

SummaryEight crew members wore an Astroskin prototype during two 30-day missions in NASA’s ground-based Human Exploration Research Analog. The team assessed wake and sleep recordings, exercise monitoring, and how easily crew could use the system.

Key FindingThe team obtained usable physiological data across 24-hour wake and sleep periods, and crew described the prototype as comfortable and easy to use. The study tested feasibility in a ground analog.

SummaryBennett and colleagues evaluated whether thermal video could recover heart rate using adaptive Eulerian Video Magnification. A seated subject wore a Hexoskin smart shirt to provide the ECG reference while light and thermal cameras recorded simultaneously. Wide-band processing identified a chest region of interest, and narrow-band thermal processing produced an oscillation at 0.9 Hz, matching the ECG frequency of about 54 beats/min. The proof of concept supports further study of contactless physiological-signal extraction but does not establish validation across people or conditions.

Key FindingThe narrow-band thermal signal and Hexoskin ECG both oscillated at 0.9 Hz, approximately 54 beats/min, and the authors reported a strong correlation. This was a seated single-subject proof of concept.

SummaryThis doctoral thesis examined whether Hexoskin respiratory measurements were stable enough for monitoring changes in lung volume. Researchers used a torso simulator to test the shirts’ response to changes in temperature and circumference. A separate experiment compared shirt measurements with spirometry during seven everyday tasks in 15 healthy adults. The human experiment did not establish accuracy in patients with COPD.

Key FindingNine of 12 Hexoskin shirts showed a linear response to changes in torso circumference on the simulator. Temperature correction reduced median drift from −34.7 to −0.5 mL/°C, identifying a source of measurement error that can be tested independently of breathing changes.

SummaryBennett and colleagues used Hexoskin to monitor ECG-derived beat intervals and heart-rate variability during 1.5- to 2.5-hour League of Legends and Overwatch sessions. The results describe 19 male players across two cohorts, while the limitations describe n=10 and two device failures, leaving an unresolved internal count discrepancy. Winning-team mean SDRR was lower, but substantial individual variability and missing high-resolution gameplay context prevented a confident link between heart-rate variability and performance.

Key FindingHexoskin HRV showed a moderate association with esports activity and a team-level correlation of r=0.78 between winning and losing teams' global HRV patterns. The winning team had lower mean SDRR, but substantial individual variability and absent high-resolution gameplay context prevented a confident link between HRV and performance.

SummaryBetson and colleagues narratively reviewed approaches for quantifying threat-or-challenge responses among undergraduate paramedicine students during high-stress clinical scenarios. The review is motivated by the mental, physical, and emotional demands of paramedicine practice. Within its survey of cardiovascular measurement options, the review identifies Hexoskin as a portable technology for monitoring cardiovascular responses.

Key FindingThe narrative review treats challenge and threat as physiologically distinguishable stress appraisals and surveys cardiovascular and hormonal measurements that could be combined in high-stress paramedicine scenarios. Hexoskin is presented as one portable cardiovascular-monitoring option, not as a device validated by a new trial in this paper.

SummaryTwenty-five healthy adults completed 30-minute pickleball-doubles and self-paced walking sessions on the same day in counterbalanced order. Hexoskin provided heart rate, step count and movement-based caloric-expenditure estimates. Pickleball produced higher heart rate, estimated caloric expenditure, perceived exertion and enjoyment, while walking produced more steps.

Key FindingMean heart rate measured with Hexoskin was nearly 14% higher and peak heart rate nearly 19% higher during pickleball doubles than walking. Estimated caloric expenditure was 36% higher during pickleball, while walking produced 54% more steps. The calorie values were movement-based predictions rather than oxygen-consumption measurements, and results from 25 healthy adults do not establish long-term health or weight-management benefits.

SummaryBirdee and colleagues conducted a 12-week randomized, single-blinded trial comparing extended-exhale and equal-ratio slow breathing. The abstract reports 100 participants, while the article body reports 99 enrolled, 97 in the intention-to-treat analysis and 94 with complete data. Participants wore Hexoskin shirts during home practice to assess adherence to the assigned breathing ratios. Slow breathing reduced PROMIS Anxiety within both groups, while neither strategy significantly changed heart-rate variability and the between-group differences were nonsignificant.

Key FindingTwelve weeks of slow breathing reduced psychological stress within both groups, but extending exhalation did not provide a significant incremental benefit over equal-ratio breathing. Neither intervention significantly changed high-frequency heart-rate variability, while Hexoskin confirmed adherence to the assigned home breathing patterns.

SummaryOne Canadian military section commander wore Hexoskin during a reserve-unit field exercise in British Columbia. The report describes 30 hours of continuous recording and a Bluetooth field test during a simulated mass-casualty event. The wearer reported comfort, no interference with duties and no apparent equipment failure during the exercise.

Key FindingThe Hexoskin garment was reported as comfortable and did not impede the wearer’s field-training tasks; no apparent failure or breakdown was observed during the exercise. It was worn over bare skin beneath a cotton T-shirt, fatigue top, rain jacket and tactical load-bearing vest with blank ammunition, without body armour or ballistic plates. This single-person training report does not establish performance across personnel, protective configurations or operational deployments.

SummaryBishnoi and Hernandez examined how heart-rate reserve was associated with prefrontal-cortex activation and gait during increasingly difficult Trail Walking Tasks. Heart-rate reserve was calculated from baseline resting heart rate and a formula-based maximum, while Hexoskin recorded heart-rate responses during the walking tasks and fNIRS measured brain activation. Higher heart-rate reserve was associated with greater prefrontal activation and shorter stride time, while stride width and stride length showed no significant heart-rate-reserve or task effects.

Key FindingIn a walking study that used Hexoskin to track heart rate during the tasks, higher heart-rate reserve was associated with greater prefrontal-cortex activation and shorter stride time during the more difficult dual-task walking conditions. Greater HbO2 activation was also associated with higher MoCA and Mini-BEST scores and faster walking, but cognitive error-rate effects were not significant.

SummaryNine law-enforcement trainees aged 21 to 31 completed one-screen and five-screen use-of-force simulator conditions. Multimodal sensors tracked cognitive-workload biomarkers, with Hexoskin supplying heart and respiratory rates. Under higher immersion, prefrontal oxygenation and heart rate increased, while respiratory rate decreased. The findings support feasibility in this preliminary setting but require confirmation in a larger sample.

Key FindingIn a study using Hexoskin for heart and respiratory rates and separate sensors for brain oxygenation, higher immersion was associated with increased oxygenation in three of four prefrontal-cortex regions and an increase in heart rate from 86.70 to 87.66 beats per minute. Respiratory rate decreased from 18.49 to 17.84 breaths per minute, a statistically significant but practically negligible change contrary to the hypothesis; bilateral Index of Cognitive Activity also changed in the opposite direction from expected. The two self-report scales showed no significant difference, and the nine-participant findings require further validation.

SummaryThe study used Hexoskin nocturnal sleep features to forecast seizure days in 78 hospitalized people with epilepsy. Support-vector-machine models normalized features to reference nights and evaluated 16-hour and 24-hour horizons with nested leave-one-patient-out cross-validation. The authors reported improvement over chance for some patients and called for longer recordings in residential settings.

Key FindingIn 78 hospitalized people with epilepsy, forecasts using Hexoskin sleep measurements beat chance for approximately 49% of patients at 16 hours and 40% at 24 hours. These percentages are the shares of patients whose forecasts outperformed chance, not the proportions of seizures predicted.

SummaryRunners used real-time Hexoskin breathing feedback during treadmill running through three visual displays: ambient Trees, directional Arrows and a continuous Path. Speeds were adjusted individually in a sport-science laboratory. Interviews, body maps and questionnaires described different ways participants perceived, adapted to and interacted with the breathing visualizations.

Key FindingSome participants reported a mismatch between their breathing and the Path triangle, while others deliberately varied their breathing to explore the Trees response. Arrow cues helped several runners understand inhale-and-exhale timing or find a more intuitive rhythm. These qualitative, self-reported experiences from a laboratory session do not establish improved breathing, running performance or training outcomes.

SummaryBläsing and colleagues simultaneously compared Hexoskin Hx1 with four other wearable ECG devices in 13 adults across four five-minute phases: rest, walking, cognitive workload and uphill walking. They assessed local ECG noise, QRS detection and RR-interval quality with morphSQ and a modified Smith-Waterman approach. Hexoskin achieved 91.33% sufficient morphological signal quality overall and a QRS positive predictive value of 0.989. Signal quality was highest at rest and during the standing two-back task and lower during level and uphill walking, showing that performance depended on the activity.

Key FindingHexoskin provided acceptable ECG performance overall, with 91.33% sufficient morphological signal quality and a QRS positive predictive value of 0.989. Sufficient signal quality was 99.61% at rest, 86.82% during level walking, 99.15% during the two-back task and 77.60% uphill. Movement produced more noise, and close shirt fit mattered because electrode displacement could reduce accuracy; Faros produced the most stable results in this simultaneous comparison.

SummaryIn an observational study, 44 healthy adults wore a Hexoskin garment while performing prompted coughs and other respiratory tasks in supine and seated positions. A random-forest model using the garment’s acceleration and respiratory-inductance features reported an F1 score of 93.0 ± 1.1%, precision of 84.2 ± 4.2% and sensitivity of 95.5 ± 1.6%. Patient monitoring was proposed as future work.

Key FindingThe cough-detection model using Hexoskin acceleration and respiration reported an F1 score of 93.0 ± 1.1%, compared with 93.1 ± 1.3% when all sensors were used. Acceleration- and respiration-based models outperformed electrocardiogram-based models for this task. The study used prompted tasks in 44 healthy adults under stationary conditions; it did not establish performance during substantial movement or in patients with spontaneous cough.

SummaryThe study used electrocardiogram signals recorded by sensors embedded in a Hexoskin shirt to classify heart-rate recovery in 38 participants after exercise. Groups were defined using the cohort median recovery value of 28 beats/min, and a support-vector classifier reported an area under the curve of 86%. The cohort-relative labels are not a clinical diagnosis or externally validated risk threshold.

Key FindingA classifier using Hexoskin ECG recordings from 38 participants achieved an area under the ROC curve of 0.86 for distinguishing faster from slower heart-rate recovery after exercise. The cutoff of 28 beats/min was the median in this group, not a validated clinical risk threshold.

SummarySchneider evaluated gait analysis from a wearable camera embedded in glasses and a Hexoskin accelerometer vest. Four participants supplied the initial video-only validation data, and five participants completed an additional treadmill collection at five speeds from 2.0 to 3.5 mph with both sensors. Computer-vision and signal-processing methods recovered walking pace in the controlled treadmill setting, and Hexoskin acceleration features were related to treadmill-derived gait speed. The author reported that one sensor could support parametric gait analysis in this setting, while retaining both could provide richer data.

Key FindingAll eight initial video segments had a camera-derived frequency peak matching the manually annotated truth peak, and sparse optical flow produced the lowest sum of squared errors among the tested computer-vision methods. Hexoskin accelerometer Y- and Z-channel features each correlated 0.954 with treadmill-derived gait speed at the 0.01 significance level, while the camera V-channel feature was not significantly correlated with true speed. Generated and reference waveforms had coherence of 1 at the gait-frequency peak for each participant, and the author concluded that either sensor alone or both together could be considered under the controlled conditions.

SummaryThis nursing field study used Hexoskin shirts to record raw electrocardiogram data for heart-rate-variability analysis during morning shifts and assessed whether the measures could help describe stressful events. The analysis also considered physical activity and the ambiguity of low-strain periods. Its findings caution against treating heart-rate-variability patterns as a direct label for breaks or stress.

Key FindingIn a field study using Hexoskin ECG, higher-stress phases were observed, but the researchers could not identify them clearly for three of the five nurses. Low activity was also an unreliable marker of breaks because it occurred during documentation, telephone calls and medication preparation. The rrHRV method was described as relatively robust to movement artifacts because it calculated values only when two adjacent R peaks were detected.

SummaryThis exploratory study used Hexoskin respiratory-inductance signals from four healthy women to examine time-frequency measures of paradoxical breathing during sitting and lying simulations with different airway resistances and a weighted vest. Outcomes included phase angle, gain and coherence. The small awake simulation was not a diagnostic study in patients with sleep apnea, asthma or obesity.

Key FindingCoherence decreased under total obstruction in some comparisons, while overall gain and phase-angle effects were inconsistent across posture and obstruction conditions. Low coherence produced missing phase and gain values, and participant responses varied substantially. The weighted vest reproduced only added thoracic and abdominal load, and the four-person healthy sample does not establish clinical detection of paradoxical breathing or obesity-related effects.

SummaryResearchers benchmarked 12 open-source QRS-detection algorithms and a downstream RR-interval editing step using single-lead Hexoskin ECG from 13 healthy adults performing several activities. The comparison used a 100 ms matching tolerance and included added baseline, muscle and electrode-motion noise. It was an algorithm benchmark rather than a clinical diagnostic or patient-outcome study.

Key FindingOn single-lead Hexoskin ECG, xqrs had the highest combined precision and sensitivity among the 12 detectors, with mean sensitivity 0.996 and precision 0.991 in the primary condition. For three participants with more false positives, RR-interval editing improved precision but reduced sensitivity. False-negative interpolation remained future work, and the 13-person healthy dataset does not establish performance in post-hospital patients.

SummaryTen male elite cyclists completed two progressive cycling tests to exhaustion 3–4 days apart. Hexoskin heart rate, breathing frequency and estimated minute ventilation were compared with laboratory measurements. Heart rate and breathing frequency showed strong agreement through most work-rate stages, while ventilation estimates were less accurate, particularly at high intensity.

Key FindingIn 10 elite male cyclists, Hexoskin heart rate correlated at least 0.98 with the laboratory reference across work rates. Breathing frequency correlated at least 0.95 through 75% of maximum work rate, but estimated minute ventilation had larger errors, particularly at high intensity.

SummaryThis exploratory study compared 56 undergraduate paramedic students completing final-term clinical assessments with 15 students completing mock classroom assessments. Hexoskin recorded heart rate, respiratory rate and heart-rate variability; questionnaires and saliva assays supplied separate psychological and biomarker measures. Valid Hexoskin streams were available for 53 participants. The groups were purposefully sampled rather than randomly assigned.

Key FindingElevated anxiety was reported in 49 of 56 assessment participants, and 33% did not pass the final-term assessment. Heart rate was higher in the assessment group before and during the activity, but not afterward; heart-rate-variability findings were mixed. The mock-assessment control could also experience stress, and the observational design does not show that anxiety caused performance differences.

SummaryThis thesis included two related exercise studies. One used eight healthy adults to evaluate a frequency-domain indicator of oxygen-uptake response kinetics; the other used 15 healthy adults to develop a random-forest oxygen-uptake model from Hexoskin-derived heart-rate and respiratory features. A separate seven-person dataset selected the tree count, and model performance was evaluated with leave-one-participant-out testing.

Key FindingAcross three exercise conditions, predicted oxygen uptake had a repeated-measures correlation of 0.911 with direct measurements and bias of 0.022 L/min. Predicted response kinetics were slower and only moderately correlated with measurements. The model required individual maximum heart rate and calibrated ventilation, limiting use outside the laboratory, and the distinct study and tuning groups must not be combined.

SummaryThis study protocol combines interviews and field observation to refine a simulation-preparedness program for emergency physicians, followed by a planned randomized trial. Participants would complete four shift recordings, with randomization after the first two. Hexoskin would measure heart rate and heart-rate variability, while post-shift anxiety would be measured separately with the State-Trait Anxiety Inventory.

Key FindingThe protocol specifies Hexoskin cardiac monitoring and would enroll 42 physicians per group and compare a three-hour CRI:SIS intervention with control using on-shift heart-rate variability and post-shift anxiety. Four shift recordings and five-minute analysis periods were specified. These are planned methods, not completed enrollment, measurements or treatment effects; the protocol does not establish reduced stress, improved preparedness or patient benefit.

SummaryEight healthy adults completed exercise in heat while wearing firefighter turnout gear under no-cooling, upper-body-cooling and whole-body-cooling conditions. A Hexoskin tank-top measured heart rate, while separate instruments measured core temperature, skin temperature and oxygen consumption. All participants completed three exercise bouts with both cooling systems, whereas none began the third no-cooling bout. Both cooling conditions reduced several measures of strain relative to no cooling, but whole-body cooling showed an additional core-temperature advantage only during the third bout.

Key FindingAll eight participants completed the exercise bouts with upper-body and whole-body cooling, while none began the third no-cooling bout. Both cooling systems reduced perceived exertion and breathing discomfort relative to no cooling, but skin-wetness responses differed: wetness rose during the first bout with no cooling and upper-body cooling, but not with whole-body cooling. The additional whole-body-cooling advantage for limiting core-temperature rise appeared only in the third bout, and applicability to operational firefighting tasks remains uncertain.

SummaryThis pilot study examined heart-rate variability after visceral osteopathic manipulations directed at pericardial attachments. Sixty-four healthy adults aged 25–35 were assigned to manipulation, simulated-manipulation or no-intervention groups. Hexoskin recorded heart rate, and five-minute segments were analyzed in time and frequency domains over several study visits.

Key FindingHeart-rate-variability measures derived from Hexoskin cardiac recordings showed mixed results. RMSSD decreased across all 64 participants from the initial to final measurement, and some frequency-domain differences appeared between manipulation and simulated-manipulation conditions. The study did not provide consistent evidence that the techniques changed heart-rate variability; the healthy 25–35-year-old sample and different assignment process for the no-intervention group limit conclusions about treatment effects.

SummaryThe study combined seven days of longitudinal Hexoskin heart-rate, respiratory and movement data with demographic measures from 43 participants. These inputs were used to estimate cardiopulmonary-exercise-test VO2max. A support-vector-regression model was evaluated with nine-fold cross-validation, and SHAP analysis described the contribution of model inputs.

Key FindingEstimates of VO2max based on longitudinal Hexoskin data and demographic measures correlated with measured VO2max (Pearson r = 0.804), with mean absolute error of 0.38 ± 0.13 L/min and mean difference of 0.038 L/min. Age ranked highest among the model inputs, followed by heart rate, height, weight and acceleration. Internal cross-validation does not establish external performance or accuracy within the study's small clinical subgroups.

SummaryThe study analyzed 304 nights from 47 people with confirmed epilepsy admitted to an inpatient monitoring unit. Hexoskin sleep measures were compared with seizures annotated from video EEG. Six patients with seizures every day or night were excluded, and nights were classified by whether a seizure occurred within 24 hours after awakening.

Key FindingLower sleep efficiency measured by Hexoskin was associated with greater odds of a seizure within 24 hours after awakening; higher wake time after sleep onset showed a similar association. Nights containing a nocturnal seizure also had lower sleep efficiency, more wake time and higher odds of a subsequent seizure. These inpatient night-level associations do not establish causality or prospective forecasting performance at home.

SummaryThe study proposed a dynamic symbolic assignment method that converts ECG RR intervals into transition features for classifying normal sinus rhythm and paroxysmal atrial fibrillation. Machine-learning performance was evaluated with public PhysioNet databases. Hexoskin and a capacitive ECG chair were used separately to demonstrate normal-rhythm transition patterns.

Key FindingOn the public databases, reported maximum F-scores reached 94.6% for AFPDB and 99.8% for AFTDB under selected settings. These were repeated cross-validation results from database signals, not Hexoskin performance measures or an independent challenge test. The garment example showed normal-rhythm patterns only and did not establish atrial-fibrillation detection accuracy or clinical validation.

SummaryThis thesis used motion-capture data from 10 university participants to test LSTM and GRU models for short-term lifting-motion forecasting and prediction of self-reported exertion. Eight participants supplied training data and two supplied test data. Hexoskin heart-rate and breathing-rate measurements were collected but explicitly excluded from the analysis.

Key FindingThe preferred motion-forecasting settings used a 100 ms horizon for LSTM and a 50 ms horizon for GRU. The fatigue models did not achieve the expected result: their predictions settled near a Borg rating of 14 and did not track higher reported exertion. Findings from two test participants in a university sample do not establish performance for experienced material-handling workers or injury prevention.

SummaryNineteen healthy adults aged 19–55 contributed laboratory or daily-life recordings for a wearable stress-prediction study. Fifteen completed laboratory testing, and eight contributed field recordings. Hexoskin heart-rate data were converted into heart-rate-variability features and paired with self-reported stress ratings for comparison of five regression models.

Key FindingIn a comparison using features derived from Hexoskin heart-rate data, the random forest produced the highest reported correct-entry rate at 86.3%, while CatBoost reached 85.3% and processed each record in about 2.02 ms. More than 70% of segments represented low stress ratings of 2–3, limiting the range used for training and evaluation. The small study does not establish clinical stress diagnosis or performance across more diverse stress levels.

SummaryEight healthy young adults wore a Hexoskin Smart Shirt during a 45-minute dry-immersion session. The session was used as a ground-based model of weightlessness. ECG-derived heart-rate variability and respiratory rate were analyzed in five-minute segments, with motion during blood-pressure measurement introducing some signal noise.

Key FindingDuring dry immersion, heart rate measured with Hexoskin fell by a further 5–10 beats/min, while SDNN and sample entropy changed significantly. Respiratory rate decreased from 16 to 13 breaths/min, but this change was not statistically significant, and several other heart-rate-variability measures did not change significantly. The small healthy cohort and ground-based analogue do not establish clinical or spaceflight performance.

SummaryResearchers developed HRVEST to process Hexoskin ECG from a clinical trial involving 81 emergency physicians and 324 work-shift recordings. The open-source workflow converted files and synchronized metadata. It then divided recordings into five-minute windows, corrected artifacts, classified signal quality and calculated heart-rate-variability measures from a 347 GB dataset.

Key FindingThe HRVEST workflow for Hexoskin cardiac recordings processed 413 ECG recordings in an average of 15 minutes 28 seconds on the reported laptop and retained 99.7% usable data after artifact correction. Run time increased roughly linearly with file count, but depended on hardware and selected options. This was a processing-methods evaluation, not validation of physiological stress measurement or a clinical outcome study.

SummaryForty-two adults with epilepsy wore a Hexoskin connected shirt during simultaneous hospital video-EEG monitoring. An XGBoost pipeline used ECG and accelerometer features to detect 66 focal-to-bilateral tonic-clonic seizures. Respiratory signals were recorded but were not included in the analysis.

Key FindingUsing Hexoskin ECG and acceleration data from 42 hospitalized adults with epilepsy, an offline model detected 56 of 66 focal-to-bilateral tonic-clonic seizures (84.8% sensitivity), with a median of 0.55 false alarms per 24 hours. The study did not test real-time detection at home.

SummaryThe study described a smartphone prototype for individualized activity and overweight-or-obesity risk profiling. It combined Hexoskin activity and vital-sign data with user-reported calorie intake and normalized selected variables using standard limits drawn from best-practice and national guidelines. The resulting weighted score was used to propose daily calorie plans.

Key FindingThe prototype combining body mass index with Hexoskin measurements assigned 50% of its overall weight to body mass index and 12.5% each to heart rate, breathing rate, minute ventilation and caloric expenditure. The authors reported training with 36 PhysioNet patient records and testing with randomly generated data. No prospective clinical evaluation, diagnostic accuracy or weight-loss benefit was reported, and future self-learning behavior was only proposed.

SummaryThis development paper described a smartphone prototype that combined body mass index, user-reported information and averaged Hexoskin heart-rate, breathing, ventilation, cadence and calorie data. Three male project students aged 21–23 initially tested one week of garment collection. The app calculated an overall score and selected activity and dietary plans.

Key FindingThe prototype, which included Hexoskin cardiac, respiratory and activity data, weighted body mass index at 50% of its overall score and heart rate, breathing rate, minute ventilation and caloric expenditure at 12.5% each. Evaluation also used synthetically generated values, which were not participant or wearer records. The paper did not report prospective clinical validation, diagnostic accuracy or weight-loss outcomes; continuous self-learning was described only as future work.

SummaryTwenty university students completed a laboratory protocol combining cognitive workload, evaluator presence and light cycling. A Hexoskin smart shirt recorded ECG, respiration and movement, while self-reports and interviews measured perceived stress. Technical issues left ECG available for 16 participants, and only cardiovascular measures were compared statistically.

Key FindingMean and maximum heart rate derived from Hexoskin ECG were higher during the stress trial than the control trial, while the heart-rate-variability measure RMSSD showed no significant trial effect. Self-reported distress, challenge and threat also differed across conditions. The preliminary student study had missing physiological records and does not establish performance outside the laboratory or in military populations discussed as a future application.

SummaryOne attending surgeon wore a Hexoskin Smart Shirt during 25 consecutive elective laparoscopic and endoscopic procedures. Time-stamped ECG-derived heart-rate variability and procedure video were divided into one-, two- and five-minute intervals. Technical events in the laparoscopic procedures were rated with the Operating Room Black Box and Generic Error Rating Tool.

Key FindingAcross intervals analyzed using one surgeon’s Hexoskin ECG recordings, surgical-event rates were 47–66% higher in higher heart-rate-variability stress quantiles than in lower quantiles; the strongest association used one-minute RMSSD intervals. Data came from one surgeon, and technical performance was assessed only for gastric bypass and sleeve gastrectomy. The observational association does not establish whether stress caused events or events increased stress.

SummaryThis applied case study followed resonant-frequency breathing training in one 45-year-old elite Olympic Trap shooter. During a simulated competition, Hexoskin respiration and acceleration were synchronized to identify shots and examine breathing between them. The athlete reported that the garment did not affect her shooting.

Key FindingThe athlete reproduced her individualized breathing pattern between shots with some variability, and heart-rate variability increased during separate paced-breathing exercises. Hexoskin ECG quality during the simulated competition was insufficient for heart-rate-variability analysis. Reported improvements in calm, control and shooting performance came from the athlete and coach; the uncontrolled single case and absent pre-intervention baseline do not establish a training effect.

SummaryIn a field study, 28 adults experienced virtual reality alone or in pairs while wearing a Hexoskin Smart Garment for heart-rate and movement measurement. The experience included a 30-minute active standing phase and an eight-minute passive seated phase. Questionnaires completed after both phases assessed presence, immersion, flow, anxiety, affect and copresence.

Key FindingIn a virtual reality study that tracked heart rate and movement with Hexoskin, paired participants reported greater copresence and more positive affect than solo participants. Presence, immersion, flow and state anxiety did not differ significantly, and neither heart rate nor physical movement changed significantly between groups. The convenience sample and field design support an experiential comparison, but they do not establish a physiological effect of shared virtual reality.

SummaryThe Canadian Bedrest Study used a two-arm randomized, non-blinded design in 23 healthy adults aged 55–65. Its 26-day schedule included 14 days of six-degree head-down bed rest and seven recovery days; half the participants received prescribed resistance and endurance exercise. Hexoskin recorded ECG, respiration and acceleration during designated 24-hour periods.

Key FindingIn the Canadian Bedrest Study, which used Hexoskin for physiological monitoring, 22 of 23 participants completed the bed-rest phase; one withdrew on day 3. Two control participants later withdrew from follow-up after atrial-fibrillation episodes that the ethics committee judged unrelated to the study. This protocol and methodology report coordinates broad physiological, cognitive and biomarker outcomes, but it does not provide comparative results establishing benefit from the exercise countermeasure, and its highly selected healthy cohort limits generalizability.

SummaryThe WE SENSE protocol describes a controlled longitudinal trial using wearable sensors around a low-grade influenza-vaccine challenge. The protocol planned to recruit 60 healthy adults, and 58 had enrolled by manuscript review. Astroskin was one of three complementary wearable platforms and the reference device for sleep, electrocardiography, and respiration monitoring. Planned monitoring covered 12 days: 7 days before and 5 days after vaccine administration. Data retrieval and sample analysis were still ongoing at protocol review, so no outcome results were reported.

Key FindingIn the WE SENSE protocol, which includes Astroskin monitoring, the primary planned outcome was within-person change in physiological and activity measures before versus after vaccination. The secondary planned outcome was a binary classifier of within-participant infection state. F1 score, accuracy, sensitivity, and specificity were planned evaluation measures rather than reported performance results.

SummaryThe WE SENSE protocol describes a controlled longitudinal trial using wearable sensors around a low-grade influenza-vaccine challenge. The protocol planned to recruit 60 healthy adults, and 58 had enrolled by manuscript review. Astroskin was one of three complementary wearable platforms and the reference device for sleep, electrocardiography, and respiration monitoring. Planned monitoring covered 12 days: 7 days before and 5 days after vaccine administration. Data retrieval and sample analysis were still ongoing at protocol review, so no outcome results were reported.

Key FindingIn the WE SENSE protocol, which includes Astroskin monitoring, the primary planned outcome was within-person change in physiological and activity measures before versus after vaccination. The secondary planned outcome was a binary classifier of within-participant infection state. F1 score, accuracy, sensitivity, and specificity were planned evaluation measures rather than reported performance results.

SummaryThe WE SENSE trial examined whether machine learning could detect within-person physiological and activity changes surrounding a live attenuated influenza vaccine challenge. A secondary binary classifier predicted an individualized systemic-inflammatory state from wearable and symptom data. Night-time and daytime Astroskin/Hexoskin features were evaluated as part of this controlled challenge. The study does not establish diagnosis of naturally acquired respiratory infection.

Key FindingAdding night-time and daytime Astroskin/Hexoskin data yielded a reported 24-hour-tolerance ROC-AUC of 0.91, specificity of 0.83 and F1 score of 0.65. The high-sensitivity threshold produced some overprediction at moderate inflammatory-response levels. These results come from a controlled healthy-adult challenge and were not externally validated for clinical diagnosis.

SummaryThree high-volume arthroplasty surgeons were monitored during conventional knee and hip arthroplasty in a prospective cohort study. A Hexoskin Smart Vest recorded cardiorespiratory measures and energy expenditure. Operative-limb laterality was not associated with significant differences in the measured variables for right- or left-handed surgeons. Findings about surgeon position came from this small specialist sample and were judged unlikely to be clinically significant.

Key FindingHexoskin cardiorespiratory and energy-expenditure variables did not differ significantly by operative-limb laterality for either right- or left-handed surgeons. During knee arthroplasty, consistently standing on the hand-dominant side was associated with higher HRV and lower minute ventilation, but the authors judged these differences unlikely to be clinically significant. Results from three high-volume surgeons do not establish effects on patient outcomes or broader surgical populations.

SummaryThis study monitored one surgeon during 40 consecutive unilateral knee arthroplasties, comparing 20 robotic-assisted with 20 conventional procedures. Hexoskin Smart Garments worn under surgical scrubs collected cardiorespiratory and caloric-expenditure measures as physiological-stress surrogates alongside wearable ergonomic measurements. Robotic cases took longer but had lower calorie expenditure per minute and lower mean heart rate. Whether these differences reduce surgeon pain or injury remains unknown.

Key FindingIn one surgeon monitored with Hexoskin, robotic-assisted knee arthroplasty took longer but had lower calorie expenditure per minute and lower mean heart rate than conventional surgery in this single-surgeon series. Robotic cases also involved less time in demanding lumbar flexion, while several neck and shoulder measures did not differ. The study did not establish that these physiological and ergonomic differences reduce musculoskeletal pain or injury.

SummaryThe study compared one surgeon during 15 robotic-assisted and 15 conventional unicompartmental knee arthroplasties. Hexoskin garments collected cardiorespiratory and caloric-expenditure data beneath surgical scrubs. Robotic cases took longer and used more calories, while HRV, breathing rate and minute ventilation did not differ significantly. The handheld device reduced neck flexion but increased shoulder abduction, so the expected overall ergonomic and physiological advantages were not realized.

Key FindingFor one surgeon wearing Hexoskin, robotic unicompartmental knee arthroplasty involved greater heart rate and total calorie expenditure than conventional instrumentation, while HRV, breathing rate and minute ventilation did not differ significantly. The handheld device reduced neck flexion but increased shoulder abduction. These findings came from one surgeon and do not establish patient outcomes or broader ergonomic benefit.

SummaryThis study compared 107 Quebec students in three flexible-seating and three fixed-seating classes. Student mental-health measures were analyzed by gender, and teacher data were summarized by group. Girls in flexible classrooms had lower mean scores across the four reported mental-health indicators, while boys had lower school-problem scores but higher scores on the other measures. Nocturnal heart-rate variability in the 45-student subsample, recorded over two consecutive nights with Hexoskin biometric vests, did not differ between seating groups for either gender. The observational classroom comparison does not establish that seating caused the reported differences.

Key FindingAcross all students, none of the four mental-health group comparisons reached the paper’s significance threshold. Gender-stratified results differed: several scores were lower for girls in flexible classrooms, whereas boys in flexible classrooms had higher internalizing, inattention/hyperactivity and emotional-symptom scores. Nocturnal heart-rate variability assessed using Hexoskin showed no seating-group difference in the 45-student subsample, and the nonrandomized classroom design limits causal interpretation.

SummaryTwelve runners completed a synchronized incremental treadmill protocol to exhaustion while wearing a Hexoskin smart shirt, RIP sensors and reference spirometry. A custom algorithm filtered and segmented the RIP data to estimate flow reversals and breathing patterns. It identified more than 99% of reference flow reversals, with pooled precision and recall of 0.998 and mean lag of 0.018 s (−0.067, 0.104). Performance was strongest for flow-reversal detection and breathing rate; other breathing-pattern components were less consistent.

Key FindingA custom algorithm for analyzing Hexoskin respiratory-band signals identified more than 99% of flow reversals, with pooled precision and recall of 0.998 and mean lag of 0.018 s (−0.067, 0.104). Breath-cycle time and breathing rate showed low bias and error, while duty cycle showed systematic error and breathing-rate variability had high relative error at low values. Results from 12 runners in a masked laboratory treadmill test do not establish field performance or accuracy for other respiratory measures.

SummaryResearchers tested real-time, step-adaptive sound guidance for locomotor-respiratory coupling in 16 novice-to-intermediate female runners. Hexoskin tracked respiration and cadence during indoor and outdoor step-rate-feedback runs followed by individualized breathing-cue trials. Guidance increased step-breath synchronization in the sequential within-subject comparison. The acute, nonrandomized proof of concept had no separate no-sound control and did not assess long-term learning or metabolic benefit.

Key FindingIn runners whose respiration and cadence were tracked with Hexoskin, step-adaptive guidance increased indoor and outdoor locomotor-respiratory attachment, although reported outdoor baseline values differ slightly between the Results and table. Outdoor breathing-rate variability and expiratory entrainment stability improved, while the overall inspiratory-stability effect was nonsignificant. The small sequential study lacked randomization, a true no-sound control and an accepted reference for overall coupling accuracy.

SummaryThis pediatric cardiology study compared a Hexoskin smart shirt and Corsano CardioWatch with 24-hour Holter ECG. The Hexoskin analysis included 36 children and found mean heart-rate accuracy of 87.4% (SD 11%). Accuracy fell from 94.9% in the first 12 hours to 80.0% in the second 12 hours and declined with greater movement and higher heart rates. A blinded cardiologist classified 31 of 36 Hexoskin rhythm recordings; five were obscured by artifacts or noise. Clinically significant brady- and tachyarrhythmias did not occur, so diagnostic accuracy for those rhythms was not established.

Key FindingAgainst 24-hour Holter ECG in 36 children, mean Hexoskin heart-rate accuracy fell from 94.9% in the first 12 hours to 80.0% in the second. A pediatric cardiologist classified 31 of 36 ECG rhythm strips from the shirt; clinically significant slow or fast arrhythmias did not occur, so the study could not assess detection of those rhythms.

SummaryForty-four young adults completed a relaxation baseline and four laboratory anxiety-eliciting tasks while Hexoskin recorded ECG and an Empatica E4 recorded electrodermal activity and skin temperature. The analysis used baseline-normalized multimodal features and Gaussian mixture models to form two- or three-level physiological response groups without using self-report as the training label. The two-cluster solution had an average silhouette score of 0.40, while the three-cluster solution scored 0.14. With leave-one-participant-out validation, SVM accuracy was 90.9% for binary cluster labels and 73.3% for three-class labels, but the authors note overlap, modest cluster separation, and difficulty distinguishing intermediate levels.

Key FindingThe two-cluster model combining Hexoskin ECG with electrodermal activity and skin temperature from a separate device showed moderate separation with an average silhouette score of 0.40, compared with 0.14 for the three-cluster solution. Under leave-one-participant-out validation, SVM accuracy was 90.9% for binary cluster labels and 73.3% for three-class labels, while the three-class psychophysiological groups were not statistically distinct.

SummaryTen healthy adults enrolled in a protocol with one ramp cycling test and two step-test visits, but only seven had paired step-test data; Hexoskin respiratory bands measured thoracic and abdominal motion. A separate turbine-based metabolic system measured tidal volume, breathing rate, minute ventilation, and gas exchange. Relative compartment contributions during repeated step exercise had ICCs of 0.87 at low, 0.89 at moderate, and 0.93 at heavy intensity. At matched minute ventilation, tidal volume and breathing rate showed no statistically significant difference between ramp and step protocols, although calibration limitations restricted ramp-test compartment comparisons to motion indices rather than absolute volumes.

Key FindingAmong the seven participants with paired step-test data, thoracic and abdominal contributions measured with Hexoskin respiratory bands to tidal volume were highly repeatable across low, moderate, and heavy step exercise, with ICCs of 0.87, 0.89, and 0.93. At matched minute ventilation, tidal volume and breathing rate showed no statistically significant difference between ramp and step protocols, but ramp-test compartment signals could not be reliably calibrated to absolute volumes and were analyzed as motion indices.

SummaryNine project-consortium participants completed 11 controlled manual-assembly sessions after a 20-minute rest, with approximately 1.0–1.3 hours of work per session. Hexoskin supplied 1 Hz heart rate and breathing rate, Xsens supplied 100 Hz motion data for posture and RULA scoring, and Borg ratings captured perceived fatigue every 5–6 minutes. The hybrid framework used a data-driven fatigue estimate to calibrate either the theoretical model’s fatigue rate or its work/rest input signal, reducing session-level mean squared error in both approaches. The small cohort, limited repeated sessions, literature-derived membership functions, and access to only two Hexoskin metrics make the findings preliminary rather than evidence of a deployable production-planning system.

Key FindingAcross 11 controlled assembly sessions from nine participants, both adaptation strategies reduced mean squared error between the initial theoretical fatigue model and the data-driven estimate. The evaluation is preliminary: Hexoskin access was limited to heart rate and breathing rate, the cohort was small, and repeated sessions within each participant were insufficient for within-person validation.

SummarySixteen healthy adults completed paced voluntary breathing at 7.5, 15, 30, 60, and 120 breaths per minute while Hexoskin chest and waist strain gauges recorded thoracic and abdominal movement. Relative phase was near in phase at preferred breathing frequency, with a mean angle of −5.58 ± 12.63°, and the absolute phase angle moved farther from in phase as breathing frequency increased. At 120 breaths per minute, seven participants remained in phase, while two showed thoracic-leading, five abdominal-leading, and two anti-phase patterns. The changes were gradual rather than an abrupt phase transition, and the harmonica task, missing trials, unmeasured respiratory volume, and limited frequency range restrict generalization.

Key FindingChest and abdominal movement measured with Hexoskin was in phase at lower and preferred breathing frequencies, then gradually shifted toward participant-specific thoracic-leading, abdominal-leading, or anti-phase patterns as breathing frequency increased. The study found no abrupt phase transition and did not measure inhaled or exhaled air volume; harmonica breathing, discarded trials, and the limited 7.5–120 bpm range constrain interpretation.

SummaryThis review examines how ambulatory ECG systems detect abnormal heartbeats and rhythms. It discusses Astroskin as an example of a garment that could combine ECG with movement and breathing signals when investigating motion artifacts.

Key FindingThe review identifies motion artifacts and scarce labeled examples of abnormal rhythms as obstacles to automated ECG analysis. It does not report accuracy for an Astroskin arrhythmia detector.

SummaryThis single-participant methods study evaluated an automatic algorithm for reducing baseline wander and motion artifacts in ambulatory electrocardiogram recordings. Astroskin and VIVALNK devices recorded lead II ECG and accelerometer signals during daily activities; the methods report one hour per device, while the results later state eight hours of acquisition. The pipeline classified usable 10-second segments, combined ECG-derived noise with accelerometer data, and selected examples showed lower mean squared error across the shown examples and higher correlation in most examples than the comparison algorithms. The findings are preliminary because only the author's data were tested, and the method could not fully reconstruct P and T waves when noise contaminated them.

Key FindingOn selected ECG segments from one participant monitored with Astroskin and VIVALNK, the proposed method had lower mean squared error than the comparison algorithms and higher correlation on most examples; processing across more than 600 ten-second segments averaged 1.5201 seconds per segment. The evaluation used recordings from one healthy 32-year-old man, so larger testing in healthy people and cardiac patients is still needed.

SummaryThis randomized laboratory study enrolled 76 varsity athletes and compared deep diaphragmatic breathing with progressive muscle relaxation after a common paced-breathing phase. Fitted Hexoskin Biometric Shirts captured ECG, respiration rate, tidal volume, heart rate, and RR intervals across baseline, intervention, and a later cognitive stress task. During the intervention, diaphragmatic breathing produced lower respiration rate, higher tidal volume, lower heart rate, and higher heart rate variability than progressive muscle relaxation; compared with paced breathing, it changed respiration rate and tidal volume but not heart rate or heart rate variability. The group differences did not persist during the cognitive challenge, and interpretation is limited by the absence of an inactive control, the single-session design, baseline heart rate variability imbalance, and unmatched breathing rates.

Key FindingIn varsity athletes monitored with Hexoskin, diaphragmatic breathing produced a lower breathing rate and higher tidal volume than paced breathing and produced lower heart rate and higher heart rate variability than progressive muscle relaxation during the intervention. These immediate differences did not carry into the cognitive stress task 10 minutes later, so the study does not establish a lasting stress-response benefit.

SummaryThis randomized study assigned 48 University of Pennsylvania undergraduates to video-guided mindful breathing or diaphragmatic breathing. Hexoskin recordings provided respiration rate, tidal volume, continuous ECG, RR intervals, and SDNN heart rate variability at baseline, during the intervention, and at a one-week follow-up. Both groups reduced respiration rate and increased tidal volume, but only the diaphragmatic-breathing group showed significant heart rate variability increases during the video and at follow-up. Seven participants did not return, technical failures caused missing data, and the small selective sample and brief intervention limit generalizability.

Key FindingIn undergraduates monitored with Hexoskin, diaphragmatic breathing increased heart rate variability from baseline, while mindful breathing did not reach statistical significance; the between-group difference was present during the video but absent one week later. A spirituality-by-condition interaction was only marginally significant at p = 0.06, so the proposed moderating role of spirituality remains tentative.

SummaryThis master's thesis developed the Boracle Middleman mobile application to collect, standardize, and store data from wearable devices. Hexoskin served as the direct Bluetooth Low Energy testbed on Android, streaming heart rate, respiration rate, step count, activity level, and cadence to local logs and a Firebase-backed pipeline. Tests covered 1-, 5-, and 15-minute sessions and separate five-minute sitting, walking, and jogging trials; log, local, and Firebase records matched, with no data loss observed. The work validates prototype data transport and storage under controlled developer testing, without a clinical cohort, physiological-accuracy comparison, completed iOS Hexoskin integration, or long-duration field validation.

Key FindingThe Android prototype maintained stable Hexoskin Bluetooth communication and one-to-one agreement among live logs, local storage, and Firebase during the reported short tests. These results support the software pipeline's data-preservation function, but they do not validate the physiological accuracy of the Hexoskin measurements or clinical performance.

SummaryThis single-blind crossover study analyzed 38 adults with sleep difficulty or insomnia disorder after one of 39 enrolled participants withdrew at baseline. Participants completed a two-day baseline and two seven-night home intervention periods using warming and sham eye masks, with separate portable EEG and iButton skin-temperature measurements and Hexoskin ECG-derived heart rate variability. Warming shortened objective sleep-onset latency relative to sham, while foot-proximal temperature gradient and normalized high-frequency heart rate variability increased relative to baseline. Foot-proximal temperature gradient was associated with objective sleep-onset latency, but the regression explained only 4.6% of its variance and did not retain the heart rate variability measures. The 8.5-minute reduction was below the cited 10-minute clinical-significance threshold, and interpretation is limited by the single-blind home design, portable EEG definition, and Kao Corporation funding.

Key FindingPeriocular warming shortened EEG-measured sleep-onset latency by 8.5 minutes relative to sham, below the cited 10-minute clinical-significance threshold. The foot-proximal skin-temperature gradient was associated with objective sleep-onset latency (R²=0.046); Hexoskin-derived normalized high-frequency heart rate variability increased relative to baseline but was not retained in the final regression model.

SummaryThis master’s thesis compared Hexoskin with a Cosmed reference system during treadmill exercise in healthy, physically active adults. The maximal exercise analysis included 28 participants, and 27 contributed to the submaximal analysis. During maximal exercise, heart rate and respiratory rate correlated strongly with the reference, while estimated ventilation correlated weakly.

Key FindingDuring maximal treadmill exercise, correlations between Hexoskin and Cosmed ranged from 0.97 to 0.99 for heart rate and 0.89 to 0.99 for respiratory rate, but only 0.16 to 0.37 for ventilation. The results distinguish tracking breathing frequency from estimating the volume of air moved.

SummaryOlschewski reanalysed a dataset from ten healthy adults to compare machine-learning approaches for estimating energy expenditure. One feature configuration, labelled “Hexoskin,” combined waist acceleration, heart rate, breathing frequency, and minute ventilation measured in the underlying multi-device dataset. Random Forest performed best for that configuration when testing on a held-out participant, while a multilayer perceptron performed best when testing on a held-out task. The thesis explores a sensor configuration rather than establishing the accuracy of recordings from a Hexoskin garment, and its limited tuning and preprocessing choices constrain interpretation.

Key FindingFor the feature subset labelled “Hexoskin,” the lowest reported RMSE was 1.26 with Random Forest in leave-one-subject-out testing and 1.35 with a multilayer perceptron in leave-one-task-out testing. These results concern a reanalysed sensor configuration and should not be presented as a direct validation of Hexoskin hardware.

SummaryFatisson investigated changes in heart rate variability around a pericardial osteopathic protocol in healthy adults. The study compared an intervention group with placebo and no-touch groups over four weekly appointments. Hexoskin recorded ECG-derived data for analysis in Kubios-HRV, alongside quality-of-life and mood questionnaires. Exploratory subgroup findings varied with initial LF/HF levels and do not establish a clinical treatment benefit.

Key FindingThe thesis used repeated wearable ECG measurements to examine physiological responses around manual therapy. Its exploratory subgroup findings require clinical validation and should not be presented as evidence that the protocol treats disease.

SummaryPion-Massicotte and colleagues developed a sleep-classification algorithm using ECG, respiratory movement, and accelerometry recorded with Hexoskin. Young healthy adults were monitored for two consecutive nights in a sleep laboratory alongside polysomnography. The corrected algorithm was evaluated on 35 recordings and achieved 77.4% agreement for wake, NREM, and REM classification. The study supports further investigation in healthy young sleepers; participant counts differ between the abstract and Methods, and the results should not be extrapolated to sleep disorders.

Key FindingAdding sleep-onset and sleep-offset corrections to a model using Hexoskin recordings increased three-state agreement with polysomnography from 70.2% to 77.4%; sleep/wake agreement reached 90.8%. The authors caution that findings from healthy young sleepers cannot be generalized to pathological conditions.

SummaryJain and colleagues developed a virtual-reality airway-training system that combined physical airway models and procedural sensors with a Hexoskin vest. Ten subjects took part in positional calibration, and seven clinicians evaluated realism and usability. The paper reports calibration and survey results but gives no physiological measurements from the vest.

Key FindingSeven clinicians evaluated the realism and usability of a virtual-reality airway-training system that included a Hexoskin vest. The paper reports participant feedback and system calibration, but no physiological results from the vest.

SummaryPriyanka and colleagues studied how dust exposure varies across construction activities and materials in a laboratory chamber. Optical particle counters measured PM1, PM2.5, and PM10 at two heights, while Hexoskin supplied inhalation-rate data for a hypothetical worker risk assessment. The results informed proposed locations for dust-control measures based on the activity and the areas of highest exposure. Field testing and assessment before and after installing those controls remain future work.

Key FindingCombining dust concentration with inhalation rate allowed the authors to model exposure for a worker in a high-exposure zone. Hexoskin contributed respiratory input to the model; it did not measure particulate matter, and the study did not test clinical health outcomes.

SummaryEmergency-department clinicians wore Hexoskin before, during, and after a simulated clinical event in a randomized comparison of blue and fluorescent debriefing light. The available thesis evidence reports a simulation-related stress response but no statistically significant difference between lighting groups in stress level or recovery duration.

Key FindingBlue debriefing light did not produce a statistically significant stress-recovery advantage over fluorescent light in the available results.

SummaryResearchers compared wearable respiratory monitors against six criteria for measuring how people inhale outside the laboratory. Hexoskin ranked highest among the devices reviewed. Even so, none met all six criteria without modification.

Key FindingHexoskin ranked as the most suitable wearable reviewed, but no available monitor met all six requirements for ambulatory inhalation topography without further development.

SummaryTen adult tobacco users enrolled in a three-day study of cigarette or hookah use in daily life. The researchers calibrated Hexoskin respiratory bands against a spirometer each day. Seven participants completed the study, yielding 4,733 breathing cycles for analysis of inhaled volume, flow and timing.

Key FindingCalibrated Hexoskin respiratory signals supported natural-environment inhalation topography across 4,733 analyzed breathing cycles.

SummaryResearchers tested Hexoskin and two other wearable respiratory monitors on a mechanical chest simulator. Over five repeated trials, the Hexoskin abdominal band produced the most linear response to chest expansion. Its signal also showed virtually no decay during simulated breath holds.

Key FindingHexoskin showed the strongest linearity and breath-hold stability among the three wearable respiratory monitors tested on the benchtop simulator.

SummaryResearchers illustrated a proposed framework for assessing astronaut countermeasures with data from 90 firefighter students in cold-stress training. The students wore Hexoskin shirts; Astroskin Bio-Monitor was discussed as a possible spaceflight application.

Key FindingThe case study shows how physiological data could inform individualized monitoring. It reports no astronaut cohort and does not test Astroskin performance in space.

SummaryIn a 12-day influenza-challenge study, 55 healthy adults wore Hexoskin or Astroskin garments during repeated three-minute stair tests and two-minute recoveries. A model using heart-rate variability during recovery predicted a blood-biomarker immune-response score with an area under the ROC curve of 0.73. Models based on measurements taken during activity performed less well.

Key FindingRecovery heart-rate variability from Hexoskin and Astroskin produced the study's strongest immune-response model, with AUC 0.73.

SummaryThis master’s thesis tested whether a classifier using Hexoskin recordings could distinguish wakefulness, non-REM sleep and REM sleep against polysomnography. Twenty-one healthy young adults participated in a two-night protocol; excluding one participant’s recordings left 40 nights for analysis. The best corrected classifier reached mean agreement of 77.54%, with a mean kappa of 0.54.

Key FindingAcross 40 nights in healthy young adults, the Hexoskin-based classifier had mean agreement of 77.54% with polysomnography for wake, non-REM and REM sleep. Mean kappa was 0.54, indicating moderate agreement beyond chance.

SummaryRobillard and colleagues described Astroskin’s design for monitoring physiology outside the laboratory. During the XPAntarctik expedition, four users wore the garment for a reported total of 27 days. Power constraints limited recording to selected times.

Key FindingAll four expedition users reported no major discomfort and found the garment easy to use. The field test assessed prolonged wear and usability; it did not establish measurement accuracy or continuous recording for 27 days.

SummaryTwenty-five emergency medicine residents wore Hexoskin during clinical simulations. Researchers calculated heart-rate variability from the ECG and compared the residents’ responses with difficulty ratings from five experts. The experts rated scenario difficulty consistently, but their ratings did not predict the residents’ physiological responses.

Key FindingIn simulations that used Hexoskin ECG to assess residents’ cardiac responses, modified Angoff ratings showed strong panel reliability but did not predict residents' heart rate variability changes during simulation.

SummaryTwenty-seven emergency medicine residents wore Hexoskin during six simulated encounters involving medical difficulty, equipment failure or interpersonal conflict. Residents reported different levels of distress depending on the stressor. Heart-rate variability fell from resting levels with all three stressors, but the differences between stressors were not significant.

Key FindingIn emergency medicine residents monitored with Hexoskin, different simulation stressors produced different distress responses, but all were associated with similar reductions in heart rate variability from rest.

SummaryTen volunteers wore Hexoskin while swallowing water, custard, apple, bread, and saliva so that thoracic and abdominal respiratory signals could be compared with video-annotated swallow timing. Swallows after food intake were distinguishable from normal breathing, but the study did not identify a universal respiratory pattern that separated all food structures.

Key FindingHexoskin respiratory signals supported swallow-event detection, while reliable classification of food structure remained unresolved.

SummaryEighteen professional caregivers performed transfer tasks using a wheelchair and a transfer-support robot while Hexoskin recorded ECG and respiration. After robot-assisted transfers, caregivers had lower heart rates and reported less physical stress. Heart-rate variability and cardiorespiratory coherence were generally higher with the robot.

Key FindingIn caregivers whose ECG and respiration were recorded with Hexoskin, using the transfer-support robot reduced several physiological and perceived indicators of caregiver physical burden compared with wheelchair transfers.

SummaryThe methods and results describe two surgeons who wore Hexoskin during 76 consecutive hip and knee replacements. The researchers compared physiological demand across the two operations. Hip replacement was associated with higher surgeon heart rate, minute ventilation and energy expenditure per unit time than knee replacement.

Key FindingTwo arthroplasty surgeons monitored with Hexoskin experienced greater cardiorespiratory strain during total hip arthroplasty than during total knee arthroplasty.

SummaryTwo surgeons wore Hexoskin during 70 primary and revision joint replacements. Revision procedures took longer and were associated with higher surgeon heart rate, stress index and energy expenditure. Heart-rate variability was lower during revisions than during primary procedures.

Key FindingThe two surgeons monitored with Hexoskin showed greater physiological demand during revision joint replacements than during primary joint replacements.

SummaryResearchers used Hexoskin as the respiratory reference for a smartphone system designed to track breathing and progress in mindfulness practice. The smartphone’s respiration estimate differed from the reference by an average absolute error of 1.6 breaths per minute. In a separate analysis, classifiers of mindfulness skill achieved F1 scores of 80% to 84%.

Key FindingThe smartphone’s respiratory-rate estimates had a mean absolute error of 1.6 breaths per minute against Hexoskin. A separate analysis classified three mindfulness skills.

SummaryFifty pre-service firefighters completed baseline and simulated search-and-rescue monitoring with Hexoskin after six additional recruits lost data to garment or battery problems. Women showed higher SDNN and lower mean heart rate than men, but the small female subgroup limits interpretation of the reported sex differences.

Key FindingHexoskin heart-rate variability measures were incorporated into a structured firefighter resilience assessment during simulated operational training.

SummaryNineteen male university esports players enrolled in a study of physiological responses during League of Legends or Overwatch sessions lasting 1.5 to 2.5 hours. Seventeen provided usable Hexoskin recordings. Heart rate, breathing rate and minute ventilation changed significantly from rest; in League of Legends, the winning team had higher values than the losing team.

Key FindingHexoskin recorded significant cardiac and respiratory changes during competitive esports play in 17 analyzed players.

SummaryResearchers used surface ECG and respiratory recordings from three dogs wearing Hexoskin to develop and test a neuromorphic pacemaker model. The hardware reproduced the target cardiac timing with an average oscillation period of 556 milliseconds. Its pacing also varied with respiratory phase.

Key FindingHexoskin recordings provided the physiological reference used to demonstrate respiration-adaptive timing in a low-power neuromorphic pacemaker model.

SummaryResearchers tested an algorithm that estimates respiration from ECG using a public benchmark recorded with Hexoskin Hx1 in 13 people. They tested real-time operation separately in five people using a different sensor. On the Hexoskin benchmark, the mean absolute error was 3.66 breaths per minute and the mean absolute percentage error was 23.69%.

Key FindingA public Hexoskin Hx1 dataset served as the benchmark for an ECG-derived respiration algorithm with 3.66 breaths-per-minute mean absolute error.

SummaryNine junior and 11 professional biathletes wore Hexoskin during roller-ski shooting sessions under increasing physical load. Breathing-pattern stability predicted prone shooting accuracy, while an inhaled pattern at shot release was associated with lower standing accuracy, particularly among junior athletes.

Key FindingBreathing stability was linked to prone shooting success, and inhaling at standing shot release was associated with poorer accuracy.

SummaryFour adults took part in a rotating-chair motion-sickness experiment on the ground. Researchers compared Astroskin heart and breathing rates with a Flexcomp system and its blood-pressure estimates with Finapres finger measurements.

Key FindingHeart rate averaged over 15-second blocks correlated strongly with Flexcomp (r = 0.989). Breathing-rate correlation was lower (r = 0.709), and Finapres blood-pressure readings were significantly higher than Astroskin estimates.

SummaryOne surgeon wore Hexoskin during 185 primary total knee replacements. A higher radiographic knee-girth ratio predicted higher surgeon heart rate, total energy expenditure and energy expenditure per minute. Patient body mass index did not correlate with the surgeon’s energy consumption.

Key FindingRadiographic knee-girth ratio predicted the surgeon’s energy demand measured with Hexoskin more effectively than patient body mass index.

SummaryResearchers tested whether breathing patterns recorded with a Hexoskin shirt could distinguish seizures from periods between seizures. An epilepsy specialist used video EEG to mark seizure onset and offset without seeing the shirt data. Across 67 seizures in 13 patients, both breathing rate and minute ventilation differed between seizure and nonseizure periods (p < 0.01).

Key FindingAmong four classifiers tested on Hexoskin respiratory recordings from 13 patients, a support vector machine best distinguished seizure from nonseizure periods, averaging 76.7% accuracy.

SummaryOne patient who later died from sudden unexpected death in epilepsy and two living controls wore Hexoskin during epilepsy monitoring. Weeks before the death, one seizure was followed by prolonged shallow breathing and a slow heart rate; most other recorded seizures were unremarkable. Separate heart-rate-variability analyses used the video-EEG system.

Key FindingHexoskin monitoring identified an unusual prolonged postictal cardiorespiratory disturbance weeks before sudden unexpected death in epilepsy in this single-patient report.

SummaryTwelve aircraft ground-deicing technicians wore Hexoskin during open-basket field work. The shirt updated heart rate every second, and 11 operators provided data suitable for analysis. Heart-rate-reserve and age-adjusted thresholds classified strain as excessive more often than the working-heart-rate method.

Key FindingUsing heart rate recorded with Hexoskin in aircraft deicing technicians, heart-rate-reserve and age-adjusted approaches detected more threshold exceedances than working heart rate during open-basket deicing work.

SummaryCan wearable signals estimate blood pressure during daily activities? In three participants, researchers trained personalized models using Astroskin ECG, forehead photoplethysmography and acceleration to estimate systolic and diastolic pressure. They evaluated the estimates separately for each participant.

Key FindingWith the Train1 setup, mean absolute error ranged from 6.3 to 11.3 mmHg for systolic pressure and from 5.2 to 9.8 mmHg for diastolic pressure. The lowest errors came from one participant.

SummaryThis study developed personalized cuffless blood-pressure models for activities of daily living in five healthy adults. Astroskin supplied ECG, forehead photoplethysmography and acceleration inputs to nonlinear autoregressive neural-network models and comparator pulse-arrival-time models. Estimates were compared with Portapres finger blood pressure during six and a half hours of data collection per participant. Two participants returned about four months later, and one was tested again at six months.

Key FindingAcross participants, blood-pressure models using Astroskin ECG, forehead photoplethysmography and acceleration produced mean absolute error that ranged from 6.21 to 7.58 mmHg for systolic pressure and from 5.03 to 6.16 mmHg for diastolic pressure, depending on the model. All models met the reported AAMI bias requirement, while only selected models met the systolic standard-deviation requirement. The personalized results from five young healthy participants do not establish population-level or clinical performance.

SummaryThis study evaluated uncertainty-aware fusion of three cuffless blood-pressure models in five healthy participants. Astroskin supplied ECG, PPG, and pulse-arrival-time inputs, while Portapres finger blood pressure served as the reference. Each participant completed 6.5 hours of monitoring. Fusion improved blood-pressure and prediction-interval precision by up to 24%, but the small healthy sample and participant-specific training limit generalizability.

Key FindingCombining blood-pressure estimates from models using Astroskin signals improved blood-pressure and prediction-interval precision by up to 24%; the result was obtained in five healthy participants after participant-specific training.

SummaryFifty-nine physicians and nurses in two emergency departments participated in a cluster-randomized sleep and stress management study that used Hexoskin ECG during work shifts. The thesis reports a 9% improvement in work-hour heart-rate variability and about 15 additional minutes of sleep per night in the intervention group, with stronger cardiac effects among caregivers with moderate burnout.

Key FindingA brief sleep and stress intervention was associated with longer sleep and improved shift heart-rate variability among emergency clinicians.

SummaryTwelve aircraft-deicing technicians wore Hexoskin during 15 hours and 52 minutes of open-basket field recording, with heart rate synchronized to video-coded work tasks. Heart rate exceeded the individual steady-state threshold for a cumulative 39 minutes, and fluid spraying, vehicle and basket movement, and control tasks were among the most demanding activities.

Key FindingTask-synchronized Hexoskin monitoring identified specific open-basket deicing activities associated with elevated cardiac workload.

SummaryResearchers compared Astroskin with reference methods for oxygen saturation, temperature, heart rate, systolic blood pressure, respiration and sleep at home. Several physiological measures showed strong agreement, but systolic blood pressure had low precision. The system detected sleep successfully, yet produced sleep stages in only 41% of attempts.

Key FindingAstroskin showed strong agreement for several ambulatory physiological measures, while blood-pressure precision and sleep-staging availability remained important limitations.

SummaryFifteen participants wore Hexoskin as the respiratory reference for LASense, a contact-free acoustic system. At four metres from the smart speaker, LASense had a median error of 0.21 breaths per minute. Errors remained low across several room and noise conditions.

Key FindingHexoskin reference measurements supported validation of contact-free respiration sensing at distances up to six metres.

SummaryNinety-three adults wore Hexoskin for seven days and reported their feelings and behavior throughout the day. Individual models identified six to 13 physiological states per person. Where the states matched interpretable patterns, selected transitions predicted higher negative affect or impulsivity.

Key FindingTransitions among individualized Hexoskin-derived physiological states were associated with near-term changes in negative affect and impulsivity.

SummaryTen participants completed respiratory-sinus-arrhythmia biofeedback training while a wrist-worn BreathCoach system was evaluated against Hexoskin ECG and respiration. BreathCoach averaged 9.6 milliseconds of inter-beat-interval error and 0.61 seconds of breathing-cycle-duration error, and the adaptive training improved several biofeedback and recovery measures.

Key FindingHexoskin reference data supported validation of wrist-based inter-beat interval and breathing estimates during adaptive biofeedback training.