The identification of biomarkers to discriminate Parkinson’s Disease from other motor diseases is crucial to provide suitable treatment to patients. This study proposes a novel approach for the classification of structural Magnetic Resonance Imaging (MRI), Dopamine Transporter scan data (DaTscan) and demographic information (age and gender) to differentiate PD patients, “Scans Without Evidence for Dopaminergic Deficit” (SWEDD) patients and healthy control subjects using Convolutional Neural Networks (CNN). In Control vs PD, the accuracy of the classifier increased by adding subject gender from 94.5% to 96.0%, while in PD vs SWEDD adding age lead to 88.7% accuracy using slices encompassing the basal ganglia. The CNN was not able to successfully discriminate SWEDD vs Control. Our results suggested that pattern changes in slices encompassing the basal ganglia and the mesencephalon are relevant biomarkers for PD suggesting that this approach may have the potential to aid in PD biomarkers detection.