Assessing gait dysfunction severity in Parkinson's Disease using 2-Stream Spatial-Temporal Neural Network.

Area of Science:

• Neurology
• Biomedical Engineering
• Artificial Intelligence

Background:

• Parkinson's Disease (PD) is a neurodegenerative disorder affecting millions globally, characterized by dopamine deficiency and significant gait impairments.
• Current diagnostic tools like MDS-UPDRS and H&Y scales are subjective, time-consuming, and limited in early detection.
• Existing sensor-based methods are often cumbersome and costly, hindering widespread accessibility.

Purpose of the Study:

• To develop a novel, accessible, and accurate method for evaluating gait dysfunction in Parkinson's Disease using video analysis.
• To introduce a 2-Stream Spatial-Temporal Neural Network (2S-STNN) for classifying PD based on gait patterns.
• To identify key gait attributes and body regions critical for PD assessment.

Main Methods:

• A hierarchical approach utilizing a 2-Stream Spatial-Temporal Neural Network (2S-STNN) was employed for PD gait analysis from video data.
• The 2S-STNN model processed spatial-temporal features from skeleton and silhouette streams.
• Saliency values were used to pinpoint critical body regions, and 21 specific gait attributes were analyzed.

Main Results:

• The 2S-STNN model achieved an 89% accuracy rate in classifying Parkinson's Disease gait dysfunction.
• The model outperformed existing state-of-the-art methods in accuracy.
• Key gait parameters including walking pace, step length, and neck forward angle showed strong correlations with PD severity.

Conclusions:

• The developed video-based 2S-STNN offers a comprehensive and convenient solution for evaluating and monitoring Parkinson's Disease gait impairments.
• This approach enhances diagnostic precision and accessibility in clinical settings.
• The identification of critical gait attributes provides valuable insights for personalized PD management.