Early Detection of Parkinson's Disease Through Hand-Drawn Pattern Analysis Using Deep Learning Techniques

Parkinson’s disease, a progressive neurodegenerative disorder affecting over 10 million people worldwide, is marked by motor impairments such as tremors and rigidity. Early diagnosis is vital for slowing progression, yet current methods often detect the disease too late. This study investigates hand-drawn spirals and waves as a novel, non-invasive tool for early detection. I applied transfer learning with CNNs to a Kaggle dataset, achieving a peak accuracy of 90% and outperforming traditional vocal-based diagnostic models. I also enhanced the model’s robustness through data preprocessing and augmentation, but challenges like the small dataset size and variability in handwriting styles limited overall performance. The results demonstrate handwriting analysis as a cost-effective, accessible diagnostic alternative. Future research should focus on expanding datasets, optimizing preprocessing, and clinical validation. This study emphasizes the potential of machine learning in developing innovative solutions for critical healthcare needs.