Histogram analysis of lipid-core plaques in coronary computed tomographic angiography: ex vivo validation against histology

Invest Radiol. 2013 Sep;48(9):646-53. doi: 10.1097/RLI.0b013e31828fdf9f.


Purpose: In coronary computed tomographic angiography (CTA), low attenuation of coronary atherosclerotic plaque is associated with lipid-rich plaques. However, an overlap in Hounsfield units (HU) between fibrous and lipid-rich plaque as well as an influence of luminal enhancement on plaque attenuation was observed and may limit accurate detection of lipid-rich plaques by CTA. We sought to determine whether the quantitative histogram analysis improves accuracy of the detection of lipid-core plaque (LCP) in ex vivo hearts by validation against histological analysis.

Materials and methods: Human donor hearts were imaged with a 64-slice computed tomographic scanner using a standard coronary CTA protocol, optical coherence tomography (OCT), a histological analysis. Lipid-core plaque was defined in the histological analysis as any fibroatheroma with a lipid/necrotic core diameter of greater than 200 μm and a circumference greater than 60 degrees as well as a cap thickness of less than 450 μm. In OCT, lipid-rich plaque was determined as a signal-poor region with diffuse borders in 2 quadrants or more. In CTA, the boundaries of the noncalcified plaque were manually traced. The absolute and relative areas of low attenuation plaque based on pixels with less than 30, less than 60, and less than 90 HU were calculated using quantitative histogram analysis.

Results: From 5 hearts, a total of 446 cross sections were coregistered between CTA and the histological analysis. Overall, 55 LCPs (12%) were identified by the histological analysis. In CTA, the absolute and relative areas of low attenuation plaque less than 30, less than 60, and less than 90 HU were 0.14 (0.31) mm2 (4.22% [9.02%]), 0.69 (0.95) mm2 (18.28% [21.22%]), and 1.35 (1.54) mm2 (35.65% [32.07%]), respectively. The low attenuation plaque area correlated significantly with histological lipid content (lipid/necrotic core size [in square millimeter] and a portion of lipid/necrotic core on the entire plaque) at all thresholds but was the strongest at less than 60 HU (r = 0.53 and r = 0.48 for the absolute and relative areas, respectively). Using a threshold of 1.0 mm2 or greater, the absolute plaque area of less than 60 HU in CTA yielded 69% sensitivity and 80% specificity to detect LCP, whereas sensitivity and specificity were 73% and 71% for using 25.0% or higher relative area less than 60 HU. The discriminatory ability of CTA for LCP was similar between the absolute and relative areas (the area under the curve, 0.744 versus 0.722; P = 0.37). Notably, the association of the low attenuation plaque area in CTA with LCP was not altered by the luminal enhancement for the relative (P = 0.48) but for the absolute measurement (P = 0.03). Similar results were achieved when validated against lipid-rich plaque by OCT in a subset of 285 cross sections.

Conclusions: In ex vivo conditions, the relative area of coronary atherosclerotic plaque less than 60 HU in CTA as derived from quantitative histogram analysis has good accuracy to detect LCP as compared with a histological analysis independent of differences in luminal contrast enhancement.

Publication types

  • Comparative Study
  • Evaluation Study
  • Research Support, Non-U.S. Gov't
  • Validation Study

MeSH terms

  • Adult
  • Algorithms*
  • Coronary Angiography / methods*
  • Coronary Artery Disease / diagnostic imaging*
  • Coronary Artery Disease / metabolism*
  • Female
  • Humans
  • Lipid Metabolism*
  • Male
  • Middle Aged
  • Numerical Analysis, Computer-Assisted
  • Radiographic Image Enhancement / methods
  • Radiographic Image Interpretation, Computer-Assisted / methods*
  • Tomography, X-Ray Computed / methods*