Parkinson’s disease (PD) is a movement-related neurodegenerative disease wherein Gray matter (GM) degeneration is well established in the literature. Accurate and timely diagnosis of PD is desired for a better life of a PD patient. The limited experts and objective assessments open horizons for developing computer-aided diagnostic systems using artificial intelligence (AI) and machine learning (ML). In this direction, the present work proposes a four-step computer-aided diagnostic (CAD) method ( \(PDRaDeX\) ) for classifying PD and Healthy controls (HC) using MRI data. The \(PDRaDeX\) pre-processes the MRI data to obtain modulated normalized GM volumes. The subtle changes in the brain introduced due to disease are captured and quantified using radiomics features. We extracted features from each brain region defined by the AAL3 atlas because extracting features at the brain level may suppress the regional changes. After that, we proposed \(CoReS\) framework utilizing consensus ranking from multiple feature selection methods. Also, \(CoReS\) ranks regions instead of features, which helps to identify the most atrophic regions. Finally, the Support Vector Machine (SVM) approach is used for classification. This study investigates the effect of gender in PD expressions, and hence, the performance of the \(PDRaDeX\)  is evaluated on two gender-specific age-matched datasets (MALES and FEMALES) curated from multiple publicly available datasets. The maximum area under curve (AUC) value of 81.7% is achieved for the MALES dataset. The regions of the Precentral Gyrus, Left Superior Frontal Gyrus, Putamen, and Thalamic Medial Dorsal Nucleus are found to be the most atrophic regions. Also, explainable AI-based analysis highlights the importance of high-order statistical features in decision-making. We believe that the \(PDRaDeX\) CAD method will motivate clinicians to work with the AI and ML community to establish a more generalized, clinically relevant, and interpretable diagnostic framework.