Abstract:
"Parkinson's Disease (PD) Micrographia poses a significant challenge in medical education and diagnosis. This document outlines the creation of the Artographia system, a novel application designed for PD Micrographia detection. Chapter 1 introduces the problem, highlighting the necessity for a robust tool in the medical field for identifying PD through handwritten images.
Chapter 4 delves into the methodology, elucidating the selection of a three-tier architecture, diverse datasets, and the implementation of a unique hybrid deep learning model. The strategy integrates pre-trained Convolutional Neural Networks (CNNs), namely InceptionV3 and MobileNetV2, utilizing transfer learning techniques. The chapter systematically discusses data preparation, model architecture, training with SGD optimizer, and deployment using TensorFlow Lite. Additionally, it outlines the development of a mobile application for PD micrographia learning.
Chapter 5 succinctly discusses initial test results, emphasizing quantitative metrics such as accuracy, F1 score, precision, recall, and loss. Insights into the hybrid model's training performance are gained through confusion matrices and visual analyses of 50 experiments. Further evaluation metrics include the model's accuracy (93.49%), F1 score (93.39%), loss (21.88%), precision (90.4%), and recall (96.58%)."
