Decision tree for severity assessment of neurodegenerative diseases using possibility approach and gait dynamics

Accurate and timely diagnosis of neurodegenerative disease (NDD) severity is crucial for clinical needs. Existing assessment methods for grading disease severity are mostly dependent on movement disorder specialist opinion resulting in subjectivity and inherent limitations. Gait instrumentation, on the other hand, can be used as a reliable tool to record various contrasting primary gait features. However, these features are agonized by individual's physical dimensions causing data dispersion. This study proposes normalized feature set, and a decision tree (DT) based model to evaluate NDD severity. The study investigates the use of raw sensor signals from foot resistive switches (FSR) to determine high-level gait features for detection of disease severity in patients suffering from neurodegenerative disorders. The methodology includes three-step DT-based approach. First, NDD patients were classified from healthy controls (HC). Disorders were categorized into Parkinson's disease (PD), Amyotrophic lateral sclerosis (ALS), and Huntington's disease (HD). Finally, disease severity was determined on clinical scale. Experimental results were established using 11,084 gait pattern recordings from NDD patients and HC. The proposed framework achieved a high coefficient of determination (R² ≈ 0.90) and low error rates with stratified 10-fold cross-validation. Comparatively, it outperformed conventional DT models. Statistical validation via the Wilcoxon signed-rank test confirmed the significance of the finding. The proposed computer aided gait analysis framework demonstrated high accuracy and reliability in diagnosing NDD severity. The accuracy and reliability of proposed framework for disease severity diagnosis is crucial for dose management and to determine the disease progress rate in NDD patients.