AJNR Am J Neuroradiol. 2026 Jun 10:ajnr.A9468. doi: 10.3174/ajnr.A9468. Online ahead of print.
ABSTRACT
BACKGROUND: Carotid artery calcification is commonly assessed on computed tomography angiography (CTA) using a threshold of ≥130 Hounsfield units (HU), yet this may overestimate its volume and merge distinct calcifications due to blooming and partial-volume effects. The aim of our study was to determine the effect of incrementally increasing segmentation thresholds on the number of clusters and calcification morphology on CTA in patients with carotid atherosclerosis.
METHODS: In a cross-sectional study of 107 carotid atherosclerotic plaques on CTA, three-dimensional calcification masks at the carotid bifurcation were segmented using incremental thresholds from 130 to 1000 HU. For each calcification cluster, volume, eccentricity, compactness, aspect ratio, and mean attenuation were quantified. Associations between features were assessed using Spearman correlation, and changes in cluster number and inter-parameter relationships across thresholds were analyzed.
RESULTS: The total number of calcification clusters increased at HU ranges between 130 (n = 149) to 450-500 HU (n = 188), reflecting improved separation of clusters at 450-500 HU, and decreased at thresholds above 500 HU as low-attenuation clusters were excluded. Inter-parameter correlations changed markedly between 130 and 250-400 HU, after which relationships stabilized, indicating more consistent morphology.
CONCLUSION: CTA segmentation HU thresholds for calcifications substantially influence the apparent number, volume, and morphological parameters of carotid calcifications. These findings suggest that 130 HU may misrepresent calcification burden, whereas thresholds of ∼300 HU may provide more accurate and better representation of the morphology of calcification clusters. Future studies examining the role of calcifications in plaque vulnerability should account for the influence of attenuation thresholds during calcification segmentation.
PMID:42270359 | DOI:10.3174/ajnr.A9468

