Artificial Intelligence in Healthcare for Early Parkinson’s Disease Diagnosis

Parkinson’s Disease (PD) is a rare, progressive disease that gets worse over time, and subsequently has no cure. It causes symptoms such as stiffness, and shaking, and leads to a loss of balance and coordination in day-to-day activities. Mobility challenges and speech difficulties exclusive to PD often complicate treatment for the disease. Timely detection of PD is of the highest importance, enabling patients to make the most of their remaining lives. The aging global population highlights the urgent need for accurate remote detection models. This research study analyzes various machine learning algorithms for determining whether a person is suffering from Parkinson’s disease or not. This paper deals with the implementation of 4 machine learning techniques, which include Random Forest, Support Vector Machine (SVM), K-Nearest Neighbors (KNN), and Logistic Regression—trained on the MDVP audio dataset obtained from UCIML, UC Irvine’s Machine Learning Repository. The results of the models are put up against each other based on precision, F score, recall, and accuracy as evaluation metrics. The model demonstrating the highest accuracy was the Random Forest classifier, achieving an accuracy of 94.87% with perfect sensitivity (recall) of $\mathbf{1. 0}$. Key features contributing to its efficacy include vocal fundamental frequency metrics such as PPE, MDVP (Hz), and spread1, which highlight the vital role of these features in disease detection. These findings support the integration of machine learning in telemedicine, offering solutions that promise to constructively contribute to the lives of patients suffering from this ailment.