Beyond electromyography: A systematic review and meta-analysis of AI enabled wearables for neuromuscular assessment in sports medicine and rehabilitation

Dariusz T. Skalski, Gdansk University of Physical Education and Sport, Faculty of Physical Culture, Gdansk, PolandFollow
Michał Spieszny, University of Physical Culture in Krakow, Krakow, PolandFollow
Łukasz Jaworski Dr, Specialized Hospital, Cardiac Surgery Department, Wejherowo, PolandFollow
Marcin Żak, Department of Sport Theory, The Jerzy Kukuczka Academy of Physical Education, Katowice, PolandFollow
Artur Terbalyan, Institute of Sport Science, The Jerzy Kukuczka Academy of Physical Education, Katowice, PolandFollow
Marlena Stawiarska, Institute of Sport Science, The Jerzy Kukuczka Academy of Physical Education, Katowice, PolandFollow
Wirginia Likus, Department of Anatomy, Faculty of Health Sciences in Katowice, Medical University of Silesia in Katowice, PolandFollow
Adam Maszczyk, 1. Institute of Sport Science, The Jerzy Kukuczka Academy of Physical Education, Katowice, Poland; 2. Gdansk University of Physical Education and Sport, Faculty of Physical Culture, Gdansk, PolandFollow

Abstract

Background: Conventional laboratory based EMG systems constrain neuromuscular assessment to controlled environments, limiting ecological validity in sport and rehabilitation. While AI enabled wearable sensor systems, including surface EMG (sEMG), high density EMG (HD-EMG), inertial measurement units (IMUs), and multimodal configurations, have expanded ambulatory monitoring capabilities, no prior meta-analysis has quantitatively synthesized their agreement with laboratory grade comparators or examined the moderating roles of sensor class and task domain. Objective: To determine the overall agreement between AI enabled wearable neuromuscular systems and laboratory benchmarks; to examine sensor class and task domain as moderators of measurement fidelity; to evaluate multimodal gain over unimodal configurations; and to identify calibration and artifact mitigation requirements for applied deployment. Methods: Following PRISMA 2020 guidelines, a multi database semantic search (PubMed/MEDLINE, Scopus, SPORTDiscus, Web of Science; 2000–2025) identified 22 eligible studies (N ≈ 340 participants) spanning diverse sensor types, task domains, and populations (healthy athletes, stroke survivors, post-arthroplasty patients, amputees). Agreement metrics (ICC, correlation coefficients, R²) were pooled via random effects models with Fisher's z transformation; sensor class and task domain were examined as moderators using pre specified subgroup analyses and meta-regression. Results: Under controlled isometric and resistance conditions, wearable systems achieved substantial to excellent agreement with laboratory comparators (pooled ICC ≈ 0.88–0.92). Agreement was significantly lower during high velocity locomotion (pooled r ≈ 0.60–0.70; Q-test, p < 0.001), confirming task domain as the primary moderator. Sensor class did not significantly moderate pooled validity (p = 0.17). Multimodal configurations consistently outperformed unimodal systems across all paired comparisons, with mean gains of approximately 4–5 percentage points in classification accuracy or correlation metrics. Conclusion: AI enabled wearable neuromuscular systems demonstrate clinical-grade validity under controlled conditions, with task domain, not sensor class, as the key determinant of measurement fidelity. Multimodal sensor fusion yields reliable incremental gains. These findings provide the first quantitative validity benchmark to guide evidence-based technology selection in sports medicine and rehabilitation.

Recommended Citation

Skalski DT, Spieszny M, Jaworski L, Zak M, Terbalyan A, Stawiarska M, Likus W, Maszczyk A. Beyond electromyography: A systematic review and meta-analysis of AI enabled wearables for neuromuscular assessment in sports medicine and rehabilitation. Balt J Health Phys Act. 2026;18(3):Article2. DOI: 10.29359/BJHPA.18.3.02