2023 年 43 巻 3 号 p. 110-115
Corticobasal syndrome (CBS) and progressive supranuclear palsy (PSP) are sporadic atypical parkinsonian disorders associated with 4-repeat tauopathies. These neurodegenerative conditions closely overlap in their clinical information, pathology, and genetic risk factors ; therefore, it is difficult to accurately diagnose CBS and PSP. Recently, an unsupervised machine-learning technique, called Subtype and Stage Inference (SuStaIn), has been proposed to reveal the data-driven disease phenotypes with distinct temporal progression patterns from widely available cross-sectional data. To clarify the differences in the temporal white matter (WM) degeneration patterns between CBS and PSP, this study applied SuStaIn for fractional anisotropy (FA) in regional WM, which was sensitive to WM degeneration, based on cross-sectional brain diffusion MRI (dMRI) data.
We obtained dMRI data from 15 healthy controls, 26 patients with CBS, and 25 patients with PSP. FA was calculated after fitting the diffusion tensor model to the corrected dMRI data for susceptibility and eddy-current induced geometric distortions and inter-volume subject motion. SuStaIn was applied to the cross-sectional regional WM tract FAs to identify both the disease subtypes and their trajectories with distinct WM degeneration patterns. To assess the performance of SuStaIn, the classification accuracy and sensitivity for CBS and PSP were calculated.
SuStaIn revealed that the CBS degeneration started from the fornix and stria terminalis (FSTs) and corpus callosum (CC), followed by the posterior corona radiata (PCR), posterior thalamic radiation (PTR), and cerebral peduncle (CP), and subsequently extended to the cingulum. Finally, it reached the superior cerebral peduncle (SCP) and corticospinal tract (CST). In contrast, the PSP degeneration started from the SCP and cingulum, followed by the CST, and subsequently extended to the FST and CC. Eventually, it reached the PCR, PTR, and CP. Accordingly, SuStaIn classified CBS and PSP with 0.863 accuracy (sensitivity : CBS, 0.885 ; PSP, 0.840). The results suggested the utility of SuStaIn for classifying patients with CBS and PSP and identifying temporal WM degeneration patterns in patients with CBS and PSP.
