Voxel-based morphometry reveals gray matter atrophy in cerebral small vessel disease: correlation with cognitive impairment.

Background: This study applied voxel-based morphometry (VBM) to investigate and compare cryptic alterations in the gray matter volume (GMV) microstructure of the brain in patients with cerebral small vessel disease (CSVD) to look for CSVD-associated cognitive impairment (CI) markers. The aim of the study was to assist clinicians in the early diagnosis of CI. Methods: A total of 52 CSVD patients with CI (CSVD-CI) and 48 CSVD patients with normal cognition (CSVD-NC) were recruited. VBM was applied to analyze three-dimensional-T1-weighted imaging (3D-T1WI) images to assess GMV in differential brain regions. Differences between groups were compared by Student's t-test, or Pearson Chi-squared test. To investigate the relationship between GMV alterations and clinical measures, bias correlation analyses were performed and to further investigate the correlation of GMV in differential brain regions with subdomains of CI. A receiver operating characteristic (ROC) curve was used to determine the ability of parameters such as modified CSVD score, medial temporal lobe atrophy (MTA), and right amygdaloid cortical volume alterations to identify CI in CSVD patients. Results: CSVD-CI patients showed that the GMV of the left lingual gyrus, dorsal nucleus of the thalamus and the right lingual gyrus, and the amygdala decreased the most significantly [family-wise error (FWE) corrected for the level of the cluster P<0.05]. Spearman's analysis showed that left lingual gyrus gray matter atrophy had the greatest correlation with the domain of attention (r=0.428, P<0.001). Right lingual gyrus gray matter atrophy had the greatest correlation with the attention domain (r=0.383, P<0.001). Volume changes in the right amygdala were mainly in delayed memory (r=0.411, P<0.001) and attention (r=0.409, P<0.001). Volume changes in the dorsal nucleus in the thalamus were mainly in orientation (r=0.323, P=0.001) and language (r=0.319, P=0.001). ROC analysis showed that the area under the ROC curve (AUC) areas of the modified CSVD total load score, MTA, GMV of the right amygdala, and the combined model were 0.699, 0.674, 0.727, and 0.801, respectively. A combined model of modified CSVD score, MTA, and right amygdala GMV predicted CI the best (AUC =0.801, sensitivity =0.713, specificity =0.889). Conclusions: The combined model of modified CSVD, MTA, and right amygdala GMV is expected to facilitate the detection of microstructural changes in CI in CSVD, thereby offering a more comprehensive information yield.