Purpose: The deformation-based morphometry (DBM) method could precisely detect the brain morphological changes, which has rarely been explored in bipolar disorder (BD) patients. This study utilized DBM to identify the structure of grey matter (GM) and white matter in BD patients, and compared DBM with traditional voxel-based morphometry (VBM) on decoding the abnormal changes within the brain, to provide new insights for the pathophysiological mechanism of BD.
Patients and methods: Brain structural changes in 67 BD patients and 70 healthy controls (HC) were analyzed using DBM and VBM. The spatial correlations of the two indicators in both hemispheres were calculated, regions with significant differences between BD and HC were analyzed by correlating with clinical variables. Furthermore, support vector machine classification algorithm was utilized to detect the capability of VBM, DBM, and fusing two indicators in diagnosing BD patients.
Results: DBM showed increased volumes in GM region in the insula and pregenual anterior cingulate cortex in BD patients. VBM showed reduction of the grey matter volume (GMV) from the inferior temporal gyrus, hippocampus, inferior frontal gyrus, olfactory cortex, fusiform gyrus, middle temporal gyrus, superior temporal gyrus, middle frontal gyrus, middle cingulate and paracingulate gyri, inferior occipital gyrus, Heschl's gyrus, and dorsolateral superior frontal gyrus. The white matter volume (WMV) from the thalamus, inferior frontal gyrus, pallidum, and anterior cingulum were decreased in BD patients. The spatial correlations of the two indicators in both hemispheres were moderately correlated. Furthermore, the highest classification accuracy of DBM-GM and GMV were 69.34% and 72.42%, respectively, which was further increased to 73.72% after fusing two indicators, indicating fusion as the superior strategy.
Conclusion: Our findings indicated structural abnormalities in multiple brain regions in BD patients using VBM and DBM, with different information obtained. Fusing DBM-GM and GMV significantly improved the classification accuracy, suggesting their potential as neuroimaging markers to assist the diagnosis of BD.
Keywords: bipolar disorder; deformation-based morphometry; machine learning; multivariate pattern analysis; voxel-based morphometry.
© 2025 Chen et al.