Introduction: For the diagnosis of Parkinson's disease (PD) and atypical parkinsonism (AP) using neuroimaging, structural measures have been largely employed since structural abnormalities are most noticeable in the diseases. Functional abnormalities have been known as well, though less clearly seen, and thus, the addition of functional measures to structural measures is expected to be more informative for the diagnosis. Here, we aimed to assess whether multimodal neuroimaging measures of structural and functional alterations could have potential for enhancing performance in diverse diagnostic classification problems. Methods: For 77 patients with PD, 86 patients with AP comprising multiple system atrophy and progressive supranuclear palsy, and 53 healthy controls (HC), structural and functional MRI data were collected. Gray matter (GM) volume was acquired as a structural measure, and GM regional homogeneity and degree centrality were acquired as functional measures. The measures were used as predictors individually or in combination in support vector machine classifiers for different problems of distinguishing between HC and each diagnostic type and between different diagnostic types. Results: In statistical comparisons of the measures, structural alterations were extensively seen in all diagnostic types, whereas functional alterations were limited to specific diagnostic types. The addition of functional measures to the structural measure generally yielded statistically significant improvements to classification accuracy, compared to the use of the structural measure alone. Conclusion: We suggest the fusion of multimodal neuroimaging measures as an effective strategy that could generally cope with diverse prediction problems of clinical concerns.
Bibliographical noteFunding Information:
This work was supported by the National Research Foundation of Korea grants funded by the Korean government (2020R1C1C1010435 to N.‐Y.S., 2019R1H1A2039678 and 2020R1I1A1A01061768 to C.P.).
All Science Journal Classification (ASJC) codes
- Behavioral Neuroscience