A novel perspective for exploring the relationship between cerebral small vessel disease and deep medullary veins with automatic segmentation.

BACKGROUND: This study aimed to establish an intelligent segmentation algorithm to count the number of deep medullary veins (DMVs) and analyze the relationship between DMVs and imaging markers of cerebral small vessel disease (CSVD).

METHODS: DMVs on magnetic resonance imaging (MRI) of patients with CSVD were counted by intelligent segmentation and manual counting. The dice coefficient and intraclass correlation coefficient (ICC) were used to evaluate their consistency and correlation. Structural MR images were used to assess imaging markers and total burden of CSVD. A multivariate linear regression model was used to evaluate the correlation between the number of DMVs counted by intelligent segmentation and imaging markers of CSVD, including white matter hyperintensities of the presumed vascular origin, lacune, perivascular spaces, cerebral microbleeds, and total CSVD burden.

RESULTS: A total of 305 patients with CSVD were enrolled. An intelligent segmentation algorithm was established to calculate the number of DMVs, and it was validated and tested. The number of DMVs counted intelligently significantly correlated with the manual counting method (r = 0.761, P< 0.001). The number of smart-counted DMVs negatively correlated with the imaging markers and total burden of CSVD (P< 0.001), and the correlation remained after adjusting for age and hypertension (P< 0.05).

CONCLUSIONS: The proposed intelligent segmentation algorithm, which was established to count DMVs, can provide objective and quantitative imaging information for the follow-up of patients with CSVD. DMVs are involved in CSVD pathogenesis and a likely new imaging marker for CSVD.