Digital Phenotyping in Physical Health: Integrating Advanced Imaging, Movement Analysis and Functional Rehabilitation

Abstract

Digital phenotyping is evolving from the collection of isolated digital measurements toward continuous, multimodal characterization of an individual's health state. In physical health and rehabilitation, this transition creates an opportunity to connect structural information from magnetic resonance imaging (MRI), computed tomography (CT), ultrasound and other imaging modalities with movement-derived measures from video, inertial measurement units (IMUs), gait systems, electromyography and consumer wearables. This review synthesizes the emerging evidence for integrating these data streams into clinically meaningful functional phenotypes. Advanced imaging can quantify tissue structure and imaging-derived features, while movement analysis captures performance, coordination, mobility and adaptation in real-world environments. Machine learning and deep learning provide mechanisms for feature extraction, temporal modelling and multimodal fusion, although clinical translation remains constrained by data heterogeneity, limited external validation, interoperability, privacy, bias and uncertain clinical utility. Recent reviews indicate rapid growth of digital phenotyping, wearable rehabilitation and AI-enabled movement analysis, but the evidence remains heterogeneous and frequently concentrated in feasibility or proof-of-concept studies. An integrated framework is therefore proposed in which imaging, movement and contextual data are converted into longitudinal functional phenotypes that support assessment, rehabilitation planning and outcome monitoring. The future direction is not replacement of clinical assessment but augmentation of clinician-led decision-making with objective, longitudinal and interpretable measurements.