GFFusion: Towards automated assessment of movement disorders from gait videos

The accurate diagnosis of movement disorders such as Early Onset Ataxia (EOA) and Developmental Coordination Disorder (DCD) presents clinical challenges due to the subtle and overlapping features of motor incoordination. The Gait Feature Fusion (GFFusion) framework aims to address these challenges by leveraging computer vision and machine learning technologies. This study introduces a method that utilizes 2D video recordings to capture the coronal plane movements of patients, facilitating the automated and objective analysis of gait characteristics. The GFFusion framework integrates deep learning models with human pose estimation technologies, such as AlphaPose, to enhance the accuracy and efficiency of analyzing and differentiating EOA and DCD from healthy controls. The application of this technology in clinical settings provides a non-invasive, low-cost, and accessible solution that could improve the management of movement disorders. This paper details the development and validation of the GFFusion framework and demonstrates its effectiveness through a dataset that includes children diagnosed with EOA and DCD, as well as healthy controls. The findings suggest that the GFFusion framework could serve as a potential tool in outpatient clinics, offering more objective assessments than traditional semi-quantitative clinical rating scales like the Scale for the Assessment and Rating of Ataxia (SARA).