JMIR Publications

ABSTRACT

Background:

Computer vision (CV) technologies are increasingly applied to gait and movement analysis to assess cognitive decline and mental health conditions. By extracting physical biomarkers such as gait and posture, CV offers a non-invasive, scalable approach for early identification of cognitive and psychosocial changes.

Objective:

This scoping review aims to explore how CV-based methods use physical biomarkers to predict cognitive and mental health outcomes.

Methods:

A scoping review was conducted in accordance with the Arksey and O’Malley framework and reported using PRISMA-ScR guidelines. Searches were performed across Medline, Embase, Web of Science, IEEE Xplore, and PubMed. Eligible studies were randomised controlled trials, cohort, or longitudinal studies involving human participants, both healthy and those with underlying pathology. Included studies used CV to analyse physical biomarkers for predicting cognitive or mental health outcomes, with objective comparators. Data were extracted on study characteristics, populations, CV methodologies, motion tasks, biomarkers, and outcomes.

Results:

Following study selection, 41 studies were selected. Most used markerless motion capture, pose estimation, or deep learning. Temporospatial features were the most commonly analysed biomarkers. Cognitive decline, depression, anxiety, and psychosocial well-being were frequently targeted outcomes. Many studies found significant links between these biomarkers and cognitive or mental health outcomes. They also proposed predictive models, most commonly classification or regression frameworks, compared to validated screening tools. These models aim to support early identification.

Conclusions:

Computer vision offers a promising approach to predicting cognitive and mental health outcomes. Future work should emphasise standardisation, clinical validation, and broader population applications.