The aim of this study was to compare delayed-phase computed tomography angiography (CTA) attenuation values with histopathology, in ability to differentiate between fibrous and lipid-rich plaques in an experimental rabbit model. Twelve atherosclerotic rabbits underwent CTA of the abdominal aorta. The scan protocol included early-phase scans (EP), delayed scans at 90 s after contrast injection (DP(90s)), delayed scans at 10 min after contrast injection (DP(10min)), and delayed scan with saline infusion (DP(Saline)). Plaque composition was analyzed by histopathology (% of lipid-rich, fibrous and macrophage areas) and CT attenuation values in Hounsfield units. Using histopathology as the reference standard (n = 119), the overall sensitivity, specificity and accuracy of 64-slice CTA for the detection of plaques was 59, 100 and 79% for the EP scans; 88, 100 and 94% for the DP(90s) scans; 81, 100 and 90% for the DP(10min) scans; and 53, 100 and 76% for the DP(Saline) scans. CT density measurements showed a substantial overlap between fibrous and lipid-rich plaques, and poor correlations with the percentage of macrophage areas in both fibrous and lipid-rich plaques (r = 0.408, and r = 0.333). In delayed-phase 64-slice CTA, DP(90s) images have the best diagnostic performance for the detection of aortic plaques.