Enhanced Detection of Parkinson's Disease using MRI Scans: A Comparative Study of Deep Learning Architectures

Parkinson's Disease is a neurodegenerative disorder that manifests with progressive motor decline and is often characterized by tremors, rigidity, and bradykinesia. Early diagnosis is crucial in effective treatment however; conventional diagnostic methods still detect disease at an advanced stage in most cases. This paper proposes a robust framework for PD classification based on the combination of Convolutional Neural Networks specifically VGG16, ResNet50, InceptionV3, Xception, and MobileNetV2. Among them, Xception has shown more accuracy in differentiating PD from healthy controls using T1-weighted MRI images from 664 healthy volunteers and 167 PD patients. Xception is superior to all other CNNs, outperforming them in the generation of feature extraction by improving the process, hence opening a potential to detect neurodegenerative diseases at an early stage and improve the treatment planning process.