Personalized Neurofeedback Teletherapy for Anxiety Using Wearable Data

Thesis Abstract

Anxiety disorders impose substantial personal and societal burdens, with traditional interventions often limited by accessibility, cost, and delays in treatment. This study proposes a novel, ICT-driven approach that integrates personalized neurofeedback teletherapy with real-time wearable physiological data to enhance anxiety management. The aim is to evaluate whether tailored neurofeedback delivered via teletherapy, informed by continuous wearable signals (heart rate variability, skin conductance, and respiration rate), yields superior reductions in anxiety symptoms compared to standard teletherapy. Specific objectives include (1) to develop an adaptive neurofeedback protocol that calibrates feedback parameters to individual autonomic profiles; (2) to assess the feasibility and acceptability of delivering neurofeedback through a secure telehealth platform; (3) to determine the relationship between changes in autonomic markers and self-reported anxiety using longitudinal data; (4) to examine moderating effects of baseline anxiety severity, trait worry, and digital literacy on treatment response; and (5) to explore potential mechanisms by which neurofeedback influences cognitive–emotional processing. A mixed-methods design will be employed. The quantitative strand will use a randomized controlled trial with 140 adults diagnosed with generalized anxiety disorder or subclinical anxiety, recruited from urban clinics and online communities. Participants will be randomly assigned to either personalized neurofeedback teletherapy (n=70) or standard teletherapy (n=70) for eight weeks. Wearable devices will continuously collect physiological signals, synchronized with teletherapy session data. Primary data will include changes in anxiety severity measured by the Generalized Anxiety Disorder-7 (GAD-7) and State-Trait Anxiety Inventory (STAI) scores at baseline, week 4, week 8, and a 3-month follow-up. Secondary measures will include cortisol salivary assays, sleep quality (Pittsburgh Sleep Quality Index), and cognitive performance (Stroop task, n-back). Wearable-derived metrics (e.g., RMSSD, LF/HF ratio, skin conductance level, respiration rate) will be used to generate real-time personalized feedback algorithms, anchored in the biopsychosocial model and supported by the neurovisceral integration framework. Analyses will involve hierarchical linear modeling to test trajectories of anxiety reduction over time and multilevel mediation to examine whether autonomic regulation mediates treatment effects. Regression analyses will identify predictors of response, including baseline severity and digital literacy. Time-series analyses (Cross-Correlation and Granger causality tests) will explore the temporal relationship between physiological changes and self-reported anxiety. Qualitative data will be collected via semi-structured interviews with a purposive subsample (n=20 per arm) at post-treatment to capture user experience, perceived usefulness, and barriers to adoption. Thematic analysis will be guided by the Technology Acceptance Model and the Self-Determination Theory to elucidate motivational processes and user engagement. Expected findings include greater reductions in GAD-7 and STAI scores in the personalized neurofeedback group compared with standard teletherapy, with effect sizes in the small-to-moderate range (Cohen’s d = 0.40–0.60) and sustained benefits at follow-up. Wearable-derived autonomic regulation is anticipated to mediate a portion of treatment effects, and higher digital literacy is expected to amplify outcomes. The study will also identify patient subgroups most likely to benefit, informing targeted implementation. The contribution to knowledge lies in integrating personalization of neurofeedback with ambient wearable data in a teletherapy context, advancing theoretical models of anxiety regulation by empirically linking autonomic markers with cognitive-emotional processes and treatment response. Practically, the research will offer a scalable, accessible intervention leveraging existing consumer and clinical wearables, with a secure telehealth infrastructure and evidence-based protocols that can be translated into routine mental health care. Limitations include potential adherence challenges, device variability, and generalizability to diverse populations. Recommendations for future work include refining the feedback algorithm with machine learning, exploring other anxiety-related phenotypes, and evaluating cost-effectiveness in real-world healthcare systems.