Objective: To evaluate the diagnostic potential of an automated diffusion imaging (DWI) analyses pipeline in separating patients with multiple system atrophy (MSA) and Parkinson’s disease (PD).

Background: Increased putaminal diffusivity has been shown to distinguish clinically diagnosed patients with MSA-P from PD patients with high diagnostic accuracy. However, a recent meta-analysis found substantial between-study variability. To overcome this limitation, further studies are needed to  develop and test harmonized imaging protocols and standardized analysis pipelines.

Method: MSA and PD patients were selected from our MRI database considering the following inclusion criteria: (1) a clinical diagnosis of probable MSA or PD at the last visit according to diagnostic criteria; (2) a clinical follow-up of ≥36 months; (3) a disease duration of ≤5 years. Exclusion criteria were dementia, extensive white matter lesions and vascular or space-occupying lesions within the cerebrum or motion artefacts on MRI. All patients underwent a 3T MRI protocol including 3D-T1 and DWI sequences. T1 images were exploited for automated brain segmentation using the FreeSurfer software package. Subsequently, DWI images were co-registered with T1 images and ADC as well as FA values were averaged within the segmented regions of interest.

Results: 31 MSA patients (18 MSA-P, 13 MSA-C) and 19 PD patients were included in the present study. There were no significant differences in demographics between the study groups [gender (p=0.882), age (p=0.118) and disease duration (p=0.787)]. MSA patients showed greater motor impairment than PD patients as indicated by higher Hoehn & Yahr stages (p < 0.001) and higher UPDRS III scores (p < 0.001)]. By applying aC4.5 classifier, increased ADC values within the middle cerebellar peduncle and the putamen were selected as the most useful measures to differentiate MSA from PD. This model classified all but one patient correctly (one MSA patient was misclassified as having PD). Cross-validation provides support for the generalizability of the model with a weighted average F-measure and a Cohen's kappa of 0.943 and 0.886, respectively.

Conclusion: (1) We successfully developed an automated processing pipeline to analyse DWI in patients with MSA and PD and (2) were able to demonstrate excellent diagnostic accuracy to differentiate MSA  from PD.

« Back to 2019 International Congress

MDS Abstracts - https://www.mdsabstracts.org/abstract/diagnostic-potential-of-automated-diffusion-imaging-for-the-differential-diagnosis-of-multiple-system-atrophy-and-parkinsons-disease/