Severity Classification of Parkinson’s Disease Based on Permutation-Variable Importance and Persistent Entropy

Parkinson’s disease (PD) is a neurodegenerative that causes chronic and progressive motor dysfunction. As PD progresses, patients show different symptoms at stages of the disease. The severity assessment inefficient subjective when it comes to artificial diagnosis. However, abnormal gait was contingent subject selection limited. Therefore, few-shot learning based on small sample sets critical solving problem insufficient data in patients. Using datasets from PhysioNet, this paper presents method permutation-variable importance (PVI) persistent entropy topological imprints, uses support vector machine (SVM) as classifier achieve classification includes following steps: (1) Take cycles, calculate characteristics each cycle. (2) Use random forest (RF) obtain leading factors differentiating levels. (3) time-delay embedding map into space, use analysis permutation homology entropy. (4) Borderline-SMOTE (BSM) balance data. (5) SVM classify samples for levels PD. An accuracy 98.08% achieved by 10-fold cross-validation, so our can be used an effective means computer-aided diagnosis PD, has important practical value.