Lama A.

Artificial Intelligence Specialization · Cohort 2024

Automated Classification of Parkinson's Gait Severity Through Activity Recognition

Parkinson's disease (PD) is a progressive neurodegenerative disorder characterized by motor symptoms, including tremors, rigidity, bradykinesia, and a particularly debilitating symptom known as freezing of gait (FoG). Accurate and early diagnosis of PD, especially the severity of gait disturbances, remains a significant clinical challenge. Traditional diagnostic methods often rely on subjective clinical assessments, which can be inconsistent and are not easily scalable. This underscores the need for objective, automated tools that can reliably assess and monitor gait disturbances in PD patients. In this report, we propose an approach to address this diagnostic gap through the use of 3D Convolutional Neural Networks (3D-CNN) for activity recognition and classification of Parkinson's gait severity. Our method leverages 3D video analysis to automatically detect and classify gait patterns associated with different stages of PD, with a particular focus on identifying the presence and severity of FoG. The 3D-CNN architecture is trained to analyze spatiotemporal features from video data, enabling a comprehensive assessment of gait disturbances. Our experimental results demonstrate that the proposed model achieves a classification accuracy of 89%, significantly outperforming existing methods. This high level of accuracy suggests that our approach could serve as a valuable tool for clinicians, offering a more objective and consistent means of assessing PD gait severity. Moreover, the scalability of this method presents opportunities for widespread adoption in clinical settings, potentially leading to improved patient outcomes through more precise monitoring and timely interventions.

Read the final report (PDF)

See it in the program report (PDF)