Medical diagnosis of parkinson disease driven by multiple preprocessing technique with scarce lee silverman voice treatment data

Abstract

Parkinson’s disease is a chronic neurodegenerative disorder characterized by the progressive deterioration of motor function that affects vocal performance and can result in noticeable disruption of vocal performance degradation. Early diagnosis of Parkinson’s disease is very crucial in preventing the disease’s progression. However, it is a complicated task for specialists or clinicians due to a wide scale of symptoms and progressive changes in disease’s symptoms over time. No standard framework exists to determine what percentage of diseases arises. This research work aims at early identification of patients with Parkinson’s disease using multiple preprocessing techniques known as Multi-Preprocessing System (MPS). The main aim of the proposed diagnostic framework is to investigate the potential of sustained vowel phonations of Lee Silverman Voice Treatment dataset which improve the classification performance. The proposed framework depends on three stages of preprocessing: (a) novel ensemble method for feature selection, (b) Discretization, and (c) Principal Component Analysis (PCA). The experimental results emphasize that the proposed MPS method provides additional support to significant reduction of cardinality and outperforms the state-of-the-art feature selection methods in terms of Accuracy, Sensitivity, Precision, and F-measure, using seven learning algorithms. The evaluation results show that proposed method using Random Forest (RF) achieves the highest performance in terms of accuracy as 94.98%, sensitivity as 93.18%, precision as 94.96%, and F-measure as 94.7%.