A Robust Parkinson's Disease Recognition from MRI Brain Images by Densely Multi-Dilated Densely Connected ResNet Model

Parkinson disease (PD) is a neurological ailment that can be treated that causes excruciating discomfort in the brain as well as problems with balance and dexterity. Physicians frequently employ experimental indicators, such as motorized and non-motor symptoms based on brain magnetic resonance imaging (MRI) pictures, to diagnose Parkinson's disease (PD) in a medical stage. Nevertheless, there are disparities in inter-variability because these medical evaluations heavily rely on skilled physicians. Thus, this work proposes a sophisticated deep learning (DL) method for precise brain MRI-based PD analysis. To improve the data eminence, Extended Unsharp Mask Filtering (Ex_UMF) is first used to the brain images. The suggested Multi-dilated Densely linked ResNet model (MD_DenseRNet) is then fed the preprocessed pictures to determine if the image is PD or normal. The Python platform processes and examines the generated method, and the simulation progresses using a publicly available PPMI database. A number of evaluation metrics, including computing time (CT), sensitivity, specificity, and precision, are looked at and linked to conventional research. The created MD_DenseRNet model achieved 98.87% accuracy, 97% precision, 98.1% recall, 96.9% specificity, and 0.15 ms CT in the experimental part.