Detecting early-stage Parkinson's disease from gait data.

Area of Science:

• Neurology
• Biomedical Engineering
• Data Science

Background:

• Parkinson's disease (PD) is a progressive neurodegenerative disorder affecting millions globally.
• Early PD symptoms are subtle and often missed, delaying diagnosis and treatment.
• Current diagnostic methods struggle with early detection of minute gait changes.

Purpose of the Study:

• To develop and validate a novel method for early detection and classification of Parkinson's disease using gait analysis.
• To address the challenge of identifying subtle gait variations in early-stage PD patients.
• To improve diagnostic accuracy and facilitate earlier intervention for PD.

Main Methods:

• Utilized dynamic time warping (DTW) to compute differences between gait cycles, capturing subtle temporal variations.
• Extracted gait features using K-means clustering on DTW-derived time-warping information.
• Employed logistic regression for classification, validated with Leave-One-Out and N-fold cross-validation on a dataset of 166 subjects (83 early PD, 10 moderate PD, 73 controls).

Main Results:

• Achieved a detection accuracy exceeding 98% for early PD symptoms.
• Demonstrated that DTW-based features are robust to variations in walking style and speed.
• Successfully classified early PD symptoms, outperforming conventional statistical features.

Conclusions:

• DTW-derived gait features offer a highly accurate and reliable method for early PD detection.
• This approach can significantly aid clinicians in diagnosing PD at its earliest stages.
• Early diagnosis through advanced gait analysis can lead to prompt treatment, potentially reducing symptom severity.