Design, implement, and evaluate a wearable for autonomic balance assessment in daily life

Thesis Abstract

This study addresses the persistent gap in continuous, real-world assessment of autonomic balance and its implications for daily functioning by developing a wearable device capable of capturing multimodal autonomic indicators during routine activities. The aim is to design, implement, and evaluate a wearable prototype that integrates heart rate variability (HRV), skin conductance (EDA), respiratory rate, and accelerometry to infer autonomic balance in daily life contexts. Specific objectives include (1) to design a compact wearable leveraging photoplethysmography (PPG) and Galvanic Skin Response (GSR) sensors with embedded processing for real-time feature extraction; (2) to implement an individualized data fusion algorithm that maps multimodal signals onto autonomic state estimates using a Bayesian hierarchical framework; (3) to evaluate the device’s validity against laboratory references and its reliability across daily activities; (4) to assess user acceptability and adherence over a 14-day field trial; and (5) to explore associations between autonomic balance indices and self-reported stress, fatigue, and mood using validated scales. The study adopts a mixed-methods research design anchored in the biopsychosocial model and theories of autonomic regulation, including the Polyvagal Theory and the Neurovisceral Integration Model, to interpret physiological signals in context. The population consists of healthy adults aged 20–40 years (N = 120) recruited from university communities, with a stratified sampling approach to ensure balanced representation by sex and activity level. A two-stage sampling method is employed an initial convenience sample for lab validation (n = 40) and a broader field sample for ecological testing (n = 80). Data collection employs a custom wrist-worn prototype collecting continuous HRV (SDNN, RMSSD, LF/HF), EDA, respiratory rate via thoracic belt, and tri-axial accelerometry at 256 Hz. Laboratory validation uses standardized autonomic challenges (orthostatic test, paced breathing, cold pressor) with simultaneous ECG to benchmark HRV-derived indices and cross-validated EDA responses against finger EDA. In the field phase, participants wear the device for 14 days in free-living conditions, completing ecological momentary assessments (EMAs) thrice daily assessing perceived stress, fatigue, and mood. Data analysis proceeds in sequential stages (i) signal preprocessing and artifact rejection using established guidelines for wearable data; (ii) feature extraction and time-window aggregation (5-minute and 30-second epochs) for HRV, EDA, respiration, and movement; (iii) data fusion and state estimation via a Bayesian hierarchical model linking multimodal features to latent autonomic balance states; (iv) validity and reliability assessment through concordance analysis with laboratory measures, intraclass correlation coefficients, and test–retest reliability across days; (v) cross-sectional and longitudinal analyses exploring correlations with EMA-based psychological measures and demographic covariates; (vi) hypothesis testing for differential autonomic responses across activities using mixed-effects modeling; and (vii) qualitative feedback analysis from user interviews to evaluate usability and perceived burden. Expected findings include strong convergent validity of wearable-derived autonomic balance indices with laboratory HRV and EDA measures, high intra-individual reliability across days, and meaningful associations between elevated sympathetic dominance and self-reported stress and fatigue. The study anticipates that the Bayesian fusion approach will achieve robust individual calibration while preserving generalizability across daily contexts. The intended contribution to knowledge lies in providing a rigorously evaluated, ecologically valid wearable system capable of continuous autonomic monitoring, advancing methodological integration of multimodal signals for autonomic inference, and offering practical insights into the dynamics of autonomic balance in real life. The main conclusion is that a well-calibrated, multimodal wearable can reliably track autonomic balance in daily life and stratify periods of heightened sympathetic activity associated with stress and fatigue, thereby informing interventions aimed at improving well-being and performance. Recommendations include refining on-device processing for real-time feedback, expanding age ranges and clinical populations, integrating contextual sensing (environmental factors and activity type), and exploring longitudinal outcomes on health-related endpoints.