Background: Artificial intelligence (AI)–driven wearable technologies are rapidly emerging in rehabilitation and functional assessment for patients with Parkinson’s disease (PD). However, evidence remains fragmented, integration into nursing practice is limited, and comprehensive synthesis is lacking. Objective: To summarize studies on AI-enabled wearable devices for PD rehabilitation and assessment, describing device types, monitored indicators, algorithms, and application characteristics, and identifying research gaps and barriers to clinical translation. Methods: Following the PRISMA-ScR guidelines, nine databases (CNKI, Wanfang, SinoMed, Cochrane Library, PubMed, Web of Science, CINAHL, Scopus, and Embase) were searched for original studies involving PD patients using AI-integrated wearable devices for rehabilitation, assessment, or monitoring. Two reviewers independently screened, extracted, and synthesized data to construct an evidence map. Results: A total of 1,402 records were initially retrieved, and 61 studies were included after deduplication and screening. Devices comprised sensor modules (wearable IMUs), smartwatches/wristbands, and smart insoles. Multi-sensor systems accounted for 77.05%, with accelerometers most common (95.08%) and signals predominantly collected passively (77.05%). Studies were mainly clinical or laboratory based, with single- or multi-session designs. Leave-one-out (47.54%) and k-fold (42.62%) cross-validation predominated, while external validation was scarce (4.92%). Sensitivity (65.57%) and accuracy (62.30%) were the most frequently reported metrics, indicating methodological and metric heterogeneity. Conclusions: AI-enabled wearables show promise for PD rehabilitation and remote assessment but remain detection-heavy with limited closed-loop implementation. Real-world, multicenter, and longitudinal evidence is sparse, and external validation and calibration are rarely performed. Future work should expand to non-motor and multimodal signals, routinely apply external validation and decision-curve analysis, and enhance standardization and interoperability. Developing closed-loop rehabilitation pathways that integrate assessment, intervention, feedback, and re-evaluation in home and community contexts to enhance clinical applicability and scalability.
Li et al. (Fri,) studied this question.