EMD-Based Method for Supervised Classification of Parkinson's Disease Patients Using Balance Control Data.

Area of Science:

• Biomedical Engineering
• Neuroscience
• Signal Processing

Background:

• Postural stability is crucial for human balance during standing and locomotion.
• Parkinson's disease (PD) significantly impairs postural control, leading to falls.
• Accurate differentiation between healthy individuals and PD patients is vital for timely intervention.

Purpose of the Study:

• To propose a novel methodology for differentiating between healthy subjects and Parkinson's disease (PD) patients.
• To leverage Empirical Mode Decomposition (EMD) for analyzing stabilometric signals.
• To assess the effectiveness of machine learning classifiers in PD detection.

Main Methods:

• Utilized Empirical Mode Decomposition (EMD) to decompose stabilometric signals into Intrinsic Mode Functions (IMFs).
• Extracted temporal and spectral parameters from signals and IMFs.
• Applied feature selection to identify the most relevant parameters.
• Employed machine learning classifiers (KNN, Decision Tree, Random Forest, SVM) for classification with 10-fold cross-validation.

Main Results:

• The Support Vector Machine (SVM) classifier achieved 92% performance.
• The Dempster-Shafer formalism method demonstrated a high accuracy of 96.51%.
• The proposed EMD-based feature extraction effectively differentiates PD patients.

Conclusions:

• The novel EMD-based methodology shows significant promise for accurate PD detection.
• Machine learning classifiers, particularly SVM and Dempster-Shafer formalism, are effective tools for this diagnostic task.
• This approach can contribute to better understanding and management of Parkinson's disease.