Objective: To identify a metabolic brain pattern specific to corticobasal degeneration (CBD) and to develop a machine learning algorithm that can differentiate CBD from other parkinsonian syndromes.

Background: Distinguishing between different neurodegenerative parkinsonisms in the early disease stages is often challenging. While disease-specific metabolic brain patterns have been identified for Parkinson’s disease (PD), multiple system atrophy (MSA), and progressive supranuclear palsy (PSP), only one group has identified a metabolic pattern specific to CBD, which is among the rarest atypical parkinsonian syndromes. Including CBD-related metabolic pattern (CBDRP) in the differential diagnosis algorithms can improve the diagnostic accuracy of the imaging-based diagnosis.

Method: 111 parkinsonian patients (14 CBD, 65 PD, 16 MSA, 16 PSP) and 14 healthy subjects (HS) underwent FDG PET imaging. The CBDRP was identified using the SSM-PCA method on CBD and HS images. The expression of CBDRP score was correlated with disease duration. The expression of CBDRP and previously identified and validated PD-, MSA-, and PSP-related patterns were calculated in all patients and z-scored according to HS. A multi-class support vector machine (SVM) classifier was used to differentiate among CBD, PD, MSA, and PSP using the leave-one-out cross-validation method (LOOCV).

Results: The CBDRP was characterized by relative hypometabolism in the caudate, thalamus, parietal cortex, and limited areas of the frontal cortex unilaterally (contralaterally to the more affected site). The expression of CBDRP correlated with disease duration (Pearson’s r=0.66, p=0.01). The CBDRP highly significantly differed from HS, PD, MSA, and PSP (p<0.0001, one-way ANOVA, post hoc Bonferroni’s test). The SVM classifier employing LOOCV validation achieved an area under the curve (AUC) of 0.90 across all diseases in the receiver operator characteristic (ROC) test. The differentiation between CBD and other diseases achieved AUC=0.94 (95%CI 0.90-0.99) with a specificity of 0.87 and a sensitivity of 0.86 at the optimal cut-off.

Conclusion: This study provides a metabolic pattern specific for CBD identified in a new population, and expands the automated machine learning algorithm to CBD. Further validation of the pattern and of the machine learning algorithm is underway.

« Back to 2023 International Congress

MDS Abstracts - https://www.mdsabstracts.org/abstract/identification-of-a-metabolic-brain-pattern-specific-to-corticobasal-degeneration-and-its-utility-in-differential-diagnosis/