Analysis of altered gait cycle duration in amyotrophic lateral sclerosis based on nonparametric probability density...

Area of Science:

• Neuroscience
• Biomechanical analysis
• Medical diagnostics

Background:

• Human locomotion is controlled by the central nervous system (CNS).
• Amyotrophic lateral sclerosis (ALS) is a neurodegenerative disease that affects motor neurons, potentially impacting gait.
• Altered gait rhythm and stride interval are potential indicators of neurological changes in ALS.

Purpose of the Study:

• To statistically analyze altered stride intervals in patients with ALS.
• To develop a machine learning model for distinguishing gait patterns between ALS patients and healthy controls.
• To evaluate the diagnostic performance of the proposed method.

Main Methods:

• Nonparametric Parzen-window approach to estimate probability density functions (PDFs) of stride intervals.
• Computation of mean left-foot stride interval and modified Kullback-Leibler divergence (MKLD) as key features.
• Classification using least squares support vector machine (LS-SVM) with Gaussian kernels.

Main Results:

• A correlation was observed between left and right foot stride intervals in ALS patients.
• The MKLD parameter was significantly different in ALS patients compared to healthy controls.
• LS-SVM achieved 82.8% accuracy and an AUC of 0.869, outperforming linear discriminant analysis.

Conclusions:

• Gait analysis, specifically stride interval patterns, can serve as a biomarker for ALS.
• The developed statistical and machine learning approach demonstrates high diagnostic potential for ALS.
• Nonlinear LS-SVM offers a superior method for classifying gait patterns in ALS.