Parkinson’s disease (PD) is a chronic neurological condition affecting millions globally, resulting from the deterioration or damage of dopamine-producing brain cells, crucial for motor function regulation. PD manifests with symptoms like impaired movement, balance issues, and posture problems. In order to decrease its progression and improve patients’ quality of life, early identification is essential. This study combines deep transfer learning with a Stochastic Gradient scheduler to present a unique method for handwriting analysis-based PD prediction. The study utilizes the NIATS datasets, comprising handwriting samples from individuals both with and without PD. Six distinct deep learning models—VGG16, VGG19, ResNet18, ResNet50, ResNet101, and ViT—are evaluated to determine their efficacy in PD prediction. Performance comparison across these models focuses on three key metrics: accuracy, precision, and F1 score. The results show that the VGG19 model has the highest average accuracy of 97.27% when using the suggested methods. This result highlights how well the Stochastic Gradient Learning Method works with deep transfer learning to improve the accuracy of PD predictions using handwriting data. This work offers important new information about how to use cutting-edge machine learning methods for PD early diagnosis and treatment, which could lead to better patient outcomes and more successful clinical interventions. This study uses hand-drawn spiral and wave patterns to investigate different deep transfer learning models for early PD detection. To reduce model overfitting, cutting-edge data augmentation methods like Aug Mix and Pix mix were used. With an accuracy of 97.27%, VGG19 outperformed the other models in the evaluation.
Mahendran et al. (Thu,) studied this question.