Deep Learning for Parkinson's Disease Detection: An Analytical Study

Parkinson's disease (PD) is a common brain disorder that affects the way people move and can cause symptoms like tremors, stiffness and slowed movement. Electroencephalogram (EEG) signals have shown promise in detecting the presence of the disease by capturing changes in activity associated with the prevailing medical condition. In our research, we conducted a comparative analysis of four deep learning algorithms - VGG 19, VGG16, DenseNet121 and DenseNet201 to identify the most effective algorithm to detect presence of the disease. The analysis consistently demonstrated that DenseNet201 outperforms other algorithms across all metrics. With an F1 score of 95.5 %, training accuracy of 99.74% and validation accuracy of 95.45%, DenseNet201 demonstrates high precision and recall rates. These findings highlight DenseNet201 as the most promising algorithm to accurately diagnose Parkinson's disease in a timely manner.