FCN-PD: An Advanced Deep Learning Framework for Parkinson's Disease Diagnosis Using MRI Data

Background/Objectives: Parkinson's disease (PD) is a progressive neurodegenerative disorder characterized by motor dysfunction, cognitive decline, and a diminished quality of life. Early and accurate diagnosis is essential for effective disease management. However, traditional diagnostic approaches, which rely on clinical observations and subjective assessments, often lead to delays and inaccuracies. This research aims to address these limitations by proposing FCN-PD, an advanced deep learning framework for accurate PD diagnosis using MRI data. Methods: The FCN-PD framework incorporates a hybrid feature extraction phase that combines EfficientNet to capture local spatial details and attention mechanisms to extract global contextual information. These features are then processed by a Fully Connected Network (FCN) for final classification. This architecture enables the model to effectively represent hierarchical features and handle high-dimensional MRI data while mitigating issues such as overfitting and feature redundancy. Results: The performance of FCN-PD was evaluated on three publicly available MRI datasets. On the PPMI dataset, it achieved an accuracy of 97.2%, outperforming traditional CNN-based models by 5.3%. On the OASIS dataset, the model achieved 95.6% accuracy, and on the MIRIAD dataset, it reached 96.8% accuracy. These results establish FCN-PD as a superior alternative to existing PD diagnostic methods. Conclusions: FCN-PD demonstrates significant improvements in diagnostic accuracy and efficiency for Parkinson's disease using MRI data. Its robust architecture effectively captures both local and global features, making it a promising tool for clinical integration and early PD detection, ultimately contributing to better patient outcomes.