Early Parkinson’s disease diagnosis using Transition Propagation Graph Neutral Network with Dynamic Hunting Leadership Optimization

Abstract

Parkinson’s disease (PD) is a neurological disorder characterized by tremors, rigidity, and impaired balance due to the degeneration of neurons. This paper proposes an Early Parkinson’s disease Diagnosis using Transition Propagation Graph Neutral Network with Dynamic Hunting Leadership Optimization algorithm using voice features/attributes (EPDD-TPGNN-DHLO) approach. First, the pre-processed voice data are analyzed using Synchrosqueezing Fractional Wavelet Transform (SFWT) to extract hand crafted features. Then, Dynamic Hunting Leadership Optimization (DHLO) algorithm is employed to feature selection for identifying the most relevant features. The Zebra Optimization Algorithm is employed to enhance the classification accuracy and optimize the weight parameters of Transition Propagation Graph Neural Networks (TPGNN). The proposed EPDD-TPGNN-DHLO method achieves 24.68% to 26.22% higher accuracy, 26.18% to 29.18% greater specificity, and 24.48% to 28.49% improved precision compared to the existing DPD-CNN-LSTM, PP-PD-ANN and EPDI-TSVS-ML models. Finally, the EPDD-TPGNN-DHLO approach demonstrates a significant improvement over the existing techniques.