A Deep Learning Framework with a Dual-View CNN to Detect Parkinson's disease Using MRI Images

Abstract

Parkinson’s Disease (PD) is a neurodegenerative disordercaused by the lack of dopamine in the brain. Magnetic Resonance Imaging (MRI) is used to assess structural and functional abnormalities in the brain. Although MRI helps to examine these abnormalities, early detection remains difficult due to single–view representation. To overcome this challenge, this paper introduces a dual-view deep learning framework called Dual-View Parkinson’s Detector (DVPD) to classify between Parkinson’s Disease (PD) and Healthy Controls (HC). The proposed method incorporates two CNNs - a Global brain structure and a focused PD-vulnerable region extracted from MRI slices, which process the global and focal representations separately. The model was trained using data augmentation, the Adam optimizer, and early stopping to minimize overfitting. Unlike prior approaches that rely on heavy and complex architectures, DVPD offers a simple, resource friendly solution. The results demonstrate that the dual-view approach improves subject-level accuracy over baseline single-view CNN models, providing an effective solution for PD classification using MRI.