The efficacy of multidetector computed tomography (MDCT) for assessing coronary plaque composition has not been fully elucidated by comparison with histological findings. This study investigated the efficacy and limitations of CT density for identifying non-calcified lipid-rich plaque compared with histopathological findings. We studied 41 target lesions treated by directional coronary atherectomy in 41 patients with coronary artery disease who had non-calcified plaques detected by 16-slice MDCT before intervention. The lesions were histopathologically classified as lipid-rich or non-lipid-rich plaques, as well as according to the presence or absence of histopathological microcalcification. The mean CT density was determined in 5 regions of interest per slice and compared among the groups. The optimum cut-off value for identifying lipid-rich plaque was determined by receiver operating characteristic (ROC) analysis using histological findings for reference. Eighteen lesions were histopathologically classified as lipid-rich (5 with microcalcification and 13 without it) and 23 lesions were non-lipid-rich (8 with microcalcification and 15 without it). The mean CT density was significantly lower for lipid-rich plaque without microcalcification compared with other plaque types (P = 0.0001). ROC analysis revealed that the optimum cut-off value for distinguishing lipid-rich from non-lipid-rich plaque without microcalcification was 50 HU (sensitivity: 92.3%; specificity: 93.3%). Histopathological microcalcification had a marked influence on the plaque CT density. Sixteen-slice MDCT can identify lipid-rich plaque by a low CT density. However, high CT density dose not exclude the possibility of lipid-rich plaque, and combined morphological assessment is necessary to differentiate plaque components.