Motion Analysis for People with Cerebral Palsy: A Vision Based Approach

Area of Science:

• Computer Vision
• Biomedical Engineering
• Rehabilitation Science

Background:

• Cerebral palsy (CP) diagnosis and severity assessment often rely on clinical observation of motor function.
• Objective, quantitative methods for analyzing movement in individuals with CP are needed.
• Facial motion analysis offers a potential avenue for non-invasive assessment.

Purpose of the Study:

• To develop a novel methodology for classifying motion patterns in individuals with cerebral palsy (CP) using RGB image sequences.
• To create and introduce a new dataset of 2D facial landmark trajectories for individuals with and without disabilities.
• To establish informative motion descriptors for assessing the degree of disability.

Main Methods:

• Utilized the Structure from Motion (SfM) framework to reconstruct 3D facial shape and motion from 2D RGB image sequences, handling occlusions.
• Developed two distinct motion descriptors: one focusing on spatial motion distribution and the other on temporal motion patterns.
• Collected and annotated a new dataset of facial landmark trajectories from RGB videos of individuals performing specific movements.

Main Results:

• The proposed motion descriptors effectively capture informative aspects of facial movement.
• Both spatial and temporal motion descriptors demonstrated significant relevance in characterizing the degree of disability in subjects.
• The methodology successfully classified individuals with and without cerebral palsy based on 2D image sequences.

Conclusions:

• The developed motion descriptors are highly informative for assessing disability severity in individuals with CP.
• The proposed methodology provides an effective, non-invasive approach for classifying cerebral palsy based on facial motion analysis.
• This work contributes a valuable dataset and computational framework for advancing research in CP assessment.