Computer-aided detection and evaluation of lipid-rich plaque on noncontrast cardiac CT.

Abstract

Noncontrast electron beam CT (EBCT) and MDCT are established for the assessment of calcified plaque, but not lipid-rich plaque. We developed software to identify lipid-rich plaque with noncontrast electron beam tomography (EBT) and MDCT. A computer algorithm was developed to automatically find contiguous lipid-rich lesions with voxel intensities below a calculated patient-specific lipid threshold. Lipid density and lipid inhomogeneity in Hounsfield units were calculated in the proximal left coronaries of three populations: 34 low-risk patients (low-risk group < 6% Framingham risk score, no calcium), 31 high-risk patients (high-risk group > 20% Framingham risk score, no calcium), and 37 patients with calcified plaque (calcium group). The mean lipid density was -19.6 +/- 3.0 (SD) H in the low-risk group, -25.3 +/- 8.2 H in the high-risk group, and -34.3 +/- 13.0 H in the calcium group (p < 0.05). The mean lipid inhomogeneity was 17.7 +/- 3.6 H in the low-risk group, 21.5 +/- 5.5 H in the high-risk group, and 29.0 +/- 7.6 H in the calcium group (p < 0.05). The mean interscan variability in lipid density and lipid inhomogeneity were 2.0 +/- 3.3 H and 2.1 +/- 3.6 H, respectively. In five patients, the locations of lipid-rich plaque correlated well with available intravascular sonography findings. Our method may be able to identify lipid-rich plaque on noncontrast cardiac CT.