The Implementation of Sensors Utilizing Artificial Intelligence for Home Monitoring of Physical Functioning, Mobility, and Rehabilitation: A Preliminary Scoping Review

Background and/or Objectives: At-home monitoring technologies allow for continuous, real-world data collection which enables proactive care and provides a more holistic view of patient function. They are valuable in managing chronic conditions, assessing fall risk, and tracking rehabilitation progress. Advances in AI are driving efforts to enhance data collection and analysis, and improve clinical implementation. The purpose of this preliminary scoping review is to assess the landscape of monitoring technologies utilizing AI for physical functioning, mobility, and rehabilitation.

Design: A scoping review was conducted using PRISMA guidelines. We searched in Medline, PubMed, Embase, and Cochrane for studies from 2015 to 2021 including terms like AI, wearable devices, mobility. Of 426 initial articles, 148 remained after abstract screening, and 18 were included after full-text review. Studies were included if they featured participant data and AI-driven sensor analysis; those with manual data collection or no participant involvement were excluded. Extracted data covered study design, participants, sensor/AI use, and clinical outcomes.

Setting : Included studies originated from North America, Europe, China, Korea, and India.

Participants : Patient populations varied in size and included patients with Parkinson's disease, multiple sclerosis, stroke, dementia, or unspecified gait instability.

Interventions: Interventions were at-home monitoring sensors/devices, comprising smartphone apps, wearable devices, dedicated sensors, or ambient home sensors.

Main Outcome Measures: Main outcome measures were study characteristics, sensor characteristics, application of monitoring, and type of machine learning utilized.

Results: Most articles were prospective studies (10/18). The majority of devices described utilized IMU sensors (14/18), deployed via smartphones (9/18) or dedicated sensors (6/18), utilized classical machine learning (12/18) to help with activity classification (8/18) or symptom detection (7/18).

Conclusions: This preliminary scoping review highlights the diverse use of at-home monitoring technologies for physical function and mobility, showing clear clinical promise. Future work will include updated literature and outcome analysis to reflect recent advances in artificial intelligence and better assess the effectiveness of these interventions.