Investigating future closed-loop virtual reality therapeutics in chronic musculoskeletal pain: Biomarkers, machine learning and immersive technologies

Tsigarides, Jordan (2026) Investigating future closed-loop virtual reality therapeutics in chronic musculoskeletal pain: Biomarkers, machine learning and immersive technologies. Doctoral thesis, University of East Anglia.

Preview

PDF Download (27MB) | Preview

Abstract

Chronic musculoskeletal pain is a major cause of disability and health service burden, yet current treatments are largely ineffective or inaccessible. Immersive virtual reality (VR) has shown potential as a tool for pain management but is typically delivered as an open-loop, non-personalised intervention. Adaptive closed-loop VR therapeutics could deliver accessible, personalised digital pain interventions, but development requires mechanistically relevant biomarkers, embedded physiological sensing, and predictive models to enable state-responsive adaptation.

This work investigated whether sensor-derived measures, analysed using statistical and machine learning approaches, can inform future closed-loop VR therapeutics for chronic musculoskeletal pain. Three empirical studies were undertaken. PainWaves applied resting-state electroencephalography (rsEEG) within a discordant co-twin design, identifying reduced frontocentral beta band power as the most consistent rsEEG feature of chronic widespread pain after partial control for shared genetic and environmental background, with weaker alpha-related findings. VIPA-EEG examined within-session rsEEG change around attention-based VR exposure in fibromyalgia, showing short-term increases in alpha and beta band power, with alpha providing the clearest residual oscillatory effect after adjustment for the aperiodic component. VRFOCUS evaluated headset-derived eye tracking during a graded VR N-back task in healthy controls and adults with chronic musculoskeletal pain. Time-series machine learning classifiers outperformed summary-feature baselines and supported subject-independent inference of task-defined cognitive load, although performance was lower and more variable in chronic pain.

Findings from this work support beta band power as a candidate mechanistic rsEEG feature, alpha as a more state-responsive rhythm around VR exposure, and embedded eye tracking as a tractable signal for cognitive load inference. They also highlight that predictive models may not generalise as reliably to clinically heterogeneous chronic pain cohorts. Overall, this work provides methodological and conceptual foundations for future personalised closed-loop VR therapeutics in chronic pain, while clarifying that rsEEG, eye tracking and machine learning serve distinct roles within that pathway.

Item Type:	Thesis (Doctoral)
Faculty \ School:	Faculty of Medicine and Health Sciences > Norwich Medical School
Depositing User:	Chris White
Date Deposited:	21 Oct 2026 13:29
Last Modified:	21 Oct 2026 13:29
URI:	https://ueaeprints.uea.ac.uk/id/eprint/104828
DOI:

Downloads

Actions (login required)

View Item