Key points are not available for this paper at this time.
Background: Physical activity plays a vital role in caring for people with multiple sclerosis (MS). Yet, challenges like fatigue, mobility impairment, fall risk, symptom fluctuation, and poor access to rehabilitation often make sustained participation difficult. Adhering to physical activity programmes is tough. Artificial intelligence (AI) may transform wearable, smartphone, clinical, and patient-reported data into interpretable information for monitoring, prediction and individualized support. Objective: Building on the identified need for individualized support, this scoping review mapped original AI-based studies relevant to physical activity, gait, mobility, ambulation, fatigue, fall risk, and real-world functioning in people with MS. Methods: The review followed PRISMA-ScR principles. Search covered records published from 1 January 2016 to 14 May 2026 across biomedical, rehabilitation, technology-oriented, trial registry, and citation-searching sources. Eligible studies included MS or MS-specific data, an explicit AI/ML or model-derived predictive analytics component, and an activity-related clinical domain. Results: The search identified 332 records/reports. After removing 127 duplicates, 205 records were screened, 45 full texts were assessed, and 21 original AI studies met the eligibility criteria. These addressed inertial sensor deep learning, wearable gait speed estimation, postural sway fall risk prediction, connected device prediction of fatigue and health state, smartphone-based ambulation characterization, treadmill or walkway gait classification, chair stand fall status prediction, daily life gait/turning fall prediction, fear-of-falling detection, sensor-derived fatigue prediction, mobile/wearable modelling of MS, and physical activity prediction. Conclusions: AI can extract clinically relevant patterns from gait, wearable, smartphone, connected device, clinical, and patient-reported data in MS. However, evidence does not yet show that AI-supported systems improve physical activity, adherence, fatigue burden, fall risk, quality of life, or functional independence. AI is therefore promising but early, and prospective, externally validated, human-supervised, fatigue-aware, safety-sensitive studies are needed to support clinically meaningful, patient-centred MS rehabilitation and activity planning.
Deligiannis et al. (Tue,) studied this question.