Parkinson Disease Screening Using UNET Neural Network and BWO Based on Hand Drawn Pattern

The majority of diseases may be detected without the assistance of a trained medical professional or verified using artificial intelligence. Parkinson's disease has become one of such diseases; as among the telltale signs of Parkinson's disease, the patient starts to lose his or her ability to write precisely. This implies that whenever a patient is requested to draw certain patterns, such as waves or spirals, the patient lacks precision and produces a disoriented drawing. Some approaches have tried their participation in the study by using these initial signs as the push to diagnose Parkinson disease. However, a randomized trial found that the quality of the Parkinson's disease diagnosis procedure still has to be improved. As a result, the dataset of sine wave and spiral drawing samples of non-patients and patients are used in this research paper to obtain relatively high precision. This dataset is utilized to discover writing patterns by the UNET deep learning model, which is subsequently employed by the Black Widow Optimization method. Finally, using the Decision Tree approach, the collected patterns are evaluated to produce the optimized solution for Parkinson's disease identification. The experimental evaluation has resulted in achieving 94.37% precision, 94.37% recall, and 94.37% of accuracy for the proposed model.