Park-Net: A Deep Model for Early Detection of Parkinson’s Disease Through Automatic Analysis of Handwriting

The Parkinson disease, a neurodegenerative disease encompasses a myriad of complex and in-curable disorders. This research addresses the considered issue with a specific focus on its early detection through the analysis of offline handwriting, a novel approach in the field of medical diagnostics. In order to achieve this groundbreaking objective, we introduce Park-Net, a meticulously designed convolutional neural network architecture. The driving force behind this study is the pressing need for early detection methods that can revolutionize the way we approach Parkinson’s disease. Timely diagnosis is essential for better patient care. Also, the well treatment planning improves the quality of life for those affected by this kind of disease. In rigorous testing against comprehensive Parkinson’s disease handwriting datasets, Park-Net shines as a beacon of innovation. The considered approach combines the Park-Net model with a semi-SVM classifier, yielded remarkable results. We achieved a balanced accuracy of 91.67% on the challenging Spiral-Meander combination of the HandPD dataset, a staggering 98.00% on the Meander task of the NewHandPD dataset, and an impressive 96.43% on the Meander task from the same dataset. These results demonstrate a potential of the approach for accurate and early Parkinson’s disease detection. To our knowledge, Park-Net stands at the forefront of Parkinson’s disease detection research, surpassing the most recent studies so far. Our achievement represents a significant leap forward in the field of medical diagnostics, offering new hope and possibilities for early intervention and improved patient outcomes in the battle against Parkinson’s disease.