An Innovative Deep-Learning Approach for Identifying Parkinson’s Disease through Spiral and Wave Sketch Patterns

Parkinson's illness affects millions of people worldwide, impacting lives in many different ways. Thus, we need proper research and well-organized treatments to tackle Parkinson's widespread impact on people's lives. Rapidly identifying PD is important for providing essential care and managing this condition. This research paper introduces an innovative approach making the use of transfer learning algorithms on wave and spiral drawings to detect Parkinson's disease. The goal with this research is to develop a method that's spot-on in identifying Parkinson's disease. The focus is on analyzing the motor symptoms because these symtoms are displayed in spiral and wave sketches which are acquired from individuals using electronic tools. To achieve this, transfer learning algorithms are used. These algorithms utilize pre-trained neural network models to extract significant features from the drawings. The main aim is to capture the complex patterns and abnormalities associated with PD. The research was started with the collection of a large dataset. The dataset comprised spiral and wave drawings from both individuals diagnosed with Parkinson's disease and healthy individuals. These drawings were annotated as well as preprocessed for easy understanding and to ensure data quality and consistency. Transfer learning techniques were then applied to extract significant features from the drawingspr. Pre-trained models such as Xception or EfficientNetB2 were used during the process. Model was trained on dataset with 3,264 patients, achieving an impressive 96.4% accuracy and a notable average recall of 97.3%. Thus, highlighting its precision and effectiveness in identifying relevant information. It also demonstrated a high average precision of 96.7% and an overall balanced outcome with an F1 score of 96.94%.