Predicting progression of cerebral small vessel disease: relevance of carotid perivascular fat density based on computed tomography angiography

Background: Symptomatic carotid lesions surrounded by perivascular fat have been found to be associated with the presence of cerebral small vessel disease (CSVD). In this study, we investigated the possible relationship of perivascular fat density (PFD) with CSVD and its progression, independent of the presence of carotid stenosis. Methods: This study retrospectively evaluated consecutive patients without carotid stenosis who underwent carotid computed tomography angiography (CTA), computed tomography perfusion (CTP), and two brain magnetic resonance imaging (MRI) scans at Zhejiang Provincial People's Hospital (hospital I) from January 2019 to March 2024. Patients were categorized into three groups: without CSVD (n=34), with CSVD without progression (n=83), and with CSVD progression (n=146) according to MRI markers of CSVD. Additionally, 65 patients (including 22 with CSVD without progression and 43 with CSVD progression) were collected from Hangzhou Traditional Chinese Medicine Hospital (hospital II) for external validation. PFD was quantified using a dedicated software. The association between perfusion status on CTP and CSVD was assessed. The associations of PFD and imaging markers with the progression of CSVD were also analyzed. Six models based on PFD, significant clinical factors, and radiomic signatures were developed and validated to predict the CSVD progression. Results: PFD values were positively associated with lacunes, cerebral microbleeds (CMBs), and white matter hyperintensities (WMH) (all P<0.05). In addition, patients with CSVD progression had higher PFD than those without [-51.38 +/- 7.35 vs. -57.19 +/- 7.31 Hounsfield unit (HU); P<0.001]. Multivariate analysis indicated that diabetes, coronary artery disease, PFD, and radiomic signatures were independent predictors of CSVD progression. Moreover, the hybrid model showed enhanced performance and yielded the highest area under the receiver operating characteristic curve (AUC) of the receiver operator characteristic curve [training: AUC =0.818, 95% confidence interval (CI): 0.758-0.876; internal validation: AUC =0.805, 95% CI: 0.690-0.908; external validation: AUC =0.807, 95% CI: 0.676-0.921]. Conclusions: This study showed that, in participants without carotid stenosis, PFD was predictive of CSVD progression, suggesting the possible involvement of the inflammation present in perivascular fat in the pathogenesis of CSVD.