Small sample learning classifies Parkinson's disease patients based on their walking behavior.

Area of Science:

• Neuroscience
• Biophysics
• Data Science

Background:

• Parkinson's disease (PD) is a progressive neurodegenerative disorder affecting gait and daily activities.
• Current diagnostic methods may lack sensitivity for early detection of subtle gait changes.

Purpose of the Study:

• To develop and validate a novel method using recurrence plot and recurrence triangle (RT) patterns for classifying PD patients and healthy controls.
• To identify specific RT patterns indicative of healthy versus pathological gait.

Main Methods:

• Analysis of a toy model (Rössler attractor) and real-world walking datasets.
• Application of recurrence plot and recurrence triangle (RT) analysis to classify gait patterns.
• Statistical analysis to correlate specific RT patterns with gait characteristics.

Main Results:

• The RT-based approach achieved nearly 100% classification accuracy for both toy models and real-world data.
• Specific RT patterns (e.g., type 13, 20, 51 for L=4) were associated with healthy, rhythmic gait.
• Distinct RT patterns (e.g., type 1, 60, 64) were linked to the irregular and slow movements characteristic of PD.

Conclusions:

• RT patterns provide interpretable features for differentiating healthy and pathological gaits.
• The RT-based analysis shows significant potential as an accurate and interpretable tool for early PD detection and diagnosis.
• This approach may pave the way for future clinical applications in Parkinson's disease management.