The research, reported in Frontiers in Cardiovascular Medicine, included consecutive patients with suspected carotid atherosclerosis who were examined at the First Affiliated Hospital of Henan Medical University in China between December 2023 and June 2025. Of 886 people screened, 119 met the eligibility criteria. Fifty-two patients were classified as symptomatic, meaning they had an ipsilateral acute ischemic stroke or transient ischemic attack within the preceding two weeks, and 67 were classified as asymptomatic. A senior neurologist established the diagnoses using symptoms, time of onset and imaging. The mean age was 64.61 years, and 78 participants were male. Symptomatic patients were more likely to have hypertension and a history of smoking, while the groups did not differ significantly in age, sex distribution, body mass index, hyperlipidemia, coronary heart disease or diabetes.

Researchers used a three-phase carotid computed tomography angiography protocol that included unenhanced images, an arterial phase and delayed imaging acquired 90 seconds after iodinated contrast administration. Artificial-intelligence-assisted flexible registration and differential subtraction were then used to create black-blood CT images. This processing suppresses the high-density contrast inside the vessel lumen while retaining delayed enhancement in the vessel wall and plaque, improving the visibility of non-calcified plaque components. A lipid-rich necrotic core was defined as occupying more than 40% of the total plaque area. In parallel, investigators manually placed regions of interest in the perivascular adipose tissue on the unenhanced images, at least 1 mm beyond the outer arterial wall and away from the lumen and nearby bone. Each measurement was repeated three times, and the maximum and mean attenuation values were recorded in Hounsfield units.

The imaging measurements proved highly reproducible. Intraclass correlation coefficients for maximum perivascular fat density were 0.937 for repeat readings by the same observer and 0.859 between observers; for mean density, the corresponding values were 0.899 and 0.842. Radiologists also had excellent agreement when deciding whether a plaque contained a lipid-rich necrotic core, with a kappa value of 0.834. This reliability is important because the proposed use of fat density depends on obtaining consistent measurements from small regions of adipose tissue. It does not, however, establish that the method will perform similarly across other hospitals, scanners or broader patient populations.

Symptomatic patients had substantially more severe and more complex plaque. Their median stenosis was 21.37%, compared with 7.69% in asymptomatic patients, and their mean plaque burden was 72.50% rather than 55.92%. Maximum plaque thickness, eccentricity and the remodeling index were also higher. Lipid-rich necrotic core plaques were present in 47 of 52 symptomatic patients, or 90.4%, but in only 11 of 67 asymptomatic patients, or 16.4%. The groups also differed in vascular geometry: symptomatic patients had a smaller carotid common-to-internal artery angle, a smaller proximal internal carotid curvature angle, a larger bifurcation angle and a higher maximum common carotid artery area ratio. These were group differences, not longitudinal measurements of how geometry changed before an event.

The clearest perivascular signal appeared on the side containing the plaque. Maximum plaque-side fat density averaged -37.95 Hounsfield units in symptomatic patients and -49.74 in asymptomatic patients, while mean density was -59.13 and -68.97, respectively. Less-negative CT attenuation indicates greater density, and the source article interprets this pattern as a possible sign of local perivascular inflammation. Contralateral measurements did not differ significantly between the groups. Within both symptomatic and asymptomatic patients, however, plaque-side density was higher than density on the opposite side, suggesting that the phenomenon was spatially associated with the diseased artery rather than a general difference in the adipose tissue of symptomatic individuals.

Perivascular fat density also rose with several measures of plaque vulnerability. Spearman correlations between maximum fat density and lipid-rich necrotic core presence, plaque burden, eccentricity, stenosis severity, remodeling index and maximum plaque thickness were all positive, with reported p-values below 0.001. The strongest correlations involved the presence of a lipid-rich necrotic core: the correlation coefficient was 0.729 for maximum density and 0.782 for mean density. The proposed biological explanation is that inflamed arterial walls can influence nearby adipose tissue through paracrine signaling, and that this interaction can alter fat composition and increase its CT attenuation. This mechanism remains a hypothesis consistent with the observations, rather than a mechanism directly demonstrated by tissue sampling in this study.

In multivariable logistic regression adjusted for clinical variables, three factors remained independently associated with recent symptomatic status: stenosis severity, with an odds ratio of 1.287 per modeled increase; presence of a lipid-rich necrotic core, with an odds ratio of 10.344; and maximum perivascular fat density, with an odds ratio of 1.196. The confidence interval for the lipid-rich core association was wide, from 1.472 to 72.679, indicating substantial uncertainty around its estimated effect size. Maximum density alone had an optimism-corrected AUC of 0.908, compared with 0.863 for lipid-rich core presence and 0.885 for stenosis. Adding either fat density or lipid-rich core presence to stenosis increased the corrected AUC to 0.971 or 0.966, respectively, but their difference was not statistically significant.

The three-factor model produced the highest internally validated AUC of 0.972, with a 95% confidence interval of 0.944 to 0.995. At its reported optimal cutoff, the model had 94.2% sensitivity and 92.5% specificity for distinguishing symptomatic from asymptomatic patients. Bootstrap comparisons showed that the combined model significantly outperformed the stenosis-only model, with an AUC difference of 0.085 and p=0.002. Calibration analyses suggested that estimated risks agreed reasonably well with observed event rates, and exploratory decision-curve analysis indicated greater net clinical benefit than the stenosis-only model across the evaluated thresholds. These performance measures are promising but remain internal to the dataset: the investigators used 1,000 bootstrap resamples rather than prospective testing or an independent external cohort.

Several limitations constrain how the results should be interpreted. The single-center design allowed selection bias, could not establish whether changes in perivascular fat preceded symptoms and cannot exclude reverse causation. Manual region-of-interest placement prevented volumetric quantification, while the sample was too small for detailed stratified analysis of pre-examination medication use. The comparison benchmark was a stenosis-only model rather than a model containing established clinical risk factors, which may have exaggerated the imaging model’s incremental value. Decision-curve findings were exploratory, residual confounding remained possible, and inclusion of only unilateral atherosclerosis limits applicability to bilateral disease. There was also no histopathological validation or prospective follow-up for future events. The results therefore support perivascular fat density as a promising candidate imaging marker of local vascular inflammation, but not as a validated standalone predictor of stroke.

Subject of Research: Association of CT-measured perivascular fat density with vulnerable carotid plaque and recent ischemic cerebrovascular events

Article Title: Association of perivascular fat density with vulnerable plaque and recent symptomatic cerebrovascular events

Article References: Yue, J.-Y., Liu, H.-P., Li, R., Ji, K., Yin, Z., Ding, S.-G., & Liang, C.-H. (2026). Association of perivascular fat density with vulnerable plaque and recent symptomatic cerebrovascular events. Frontiers in Cardiovascular Medicine, 13, Article 1965601. https://doi.org/10.3389/fcvm.2026.1965601

Image Credits: AI Generated

DOI: 10.3389/fcvm.2026.1965601

Keywords: perivascular fat density, carotid atherosclerosis, vulnerable plaque, lipid-rich necrotic core, black-blood CT, ischemic stroke, transient ischemic attack, cerebrovascular imaging, plaque inflammation, Association, perivascular, density