Classification of Parkinson's disease with and without dopaminergic deficiency based on non-motor symptoms and structural neuroimaging.

The presence of non-motor symptoms (NMS) such as olfactive deficit or neuropsychiatric symptoms has been associated with the diagnosis of Parkinson's Disease (PD). NMS are also associated with different brain structural features underlying distinctive processes in PD. NMS has been poorly studied in patients with a PD-like clinical profile, showing Scans Without Evidence of Dopaminergic Deficit (SWEDD). This study proposes to compare classification models differentiating PD, SWEDD and Healthy Controls (HC) based on NMS and neurostructural factors. 683 participants (382 PD diagnosed in the last 2 years, 48 with SWEDD, 170 HC) from the PPMI dataset were compared based on available assessments. Each participant underwent an olfactive, neuropsychiatric and sleep assessment, and a 3T MRI. Brain volumes were extracted and standardized from each MRI. Classifications were based on logistic regressions using 5-fold cross-validation models combining different NMS and MRI data and determining their involvement in differentiation between patient subgroups (PD vs. SWEDD) or between patients and HC. NMS were significant factors in PD vs. SWEDD, PD vs. HC and SWEDD vs. HC classifiers, when considered alone or in combination with MRI data. No classification models were significantly different from chance based-on MRI, nor more accurate combining NMS and MRI when compared with models based on NMS only. These results highlight the importance of NMS in differentiating between PD and SWEDD, PD and HC, SWEDD and HC. However, classical imaging data such as cortical and subcortical volumetry seems insufficient to improve these classifications. Other imaging features such as connectivity could also be studied.