The aim of this study was to validate the accuracy of a commercially available intravascular ultrasound system in diagnosing plaque composition in human coronary arteries. Thirty-five coronary arteries of 18 human autopsy hearts were perfused with NaCl under a pressure of 100 mmHg and examined using a Diasonics ultrasound system and 4.8 F 20 MHz catheters. An ultrasound diagnosis was made of 139 coronary sections using previously published standard criteria and compared with histologic findings. In addition, the influence of the histologic pattern of lipid and calcific deposits on the accuracy of the ultrasound diagnosis was evaluated. Of the 25 sections with a histologically normal intima, 14 (56%) were correctly identified by ultrasound, whereas fibrotic thickening was diagnosed in the remaining 11 sections. There were 114 plaques by histology which were correctly visualized by ultrasound as plaques in all instances. Plaque calcification was correctly diagnosed in 54 of 63 (86%) sections, but massive calcifications were more reliably identified by ultrasound than small speckled calcifications (43/44 = 98% vs 11/19 = 58%, p < 0.001). Fibrosis was present in all 114 plaques and was correctly visualized by ultrasound in all instances. When lipid was diagnosed by ultrasound as a homogeneous zone of low signal intensity within a fibrous plaque as suggested in the literature, lipid accumulations were identified with a sensitivity of 26% (16/62) and a specificity of 92% (71/77). When lipids were diagnosed if more than a quarter of the plaque area showed lower signal intensity than the tissue surrounding the vessel, the sensitivity of ultrasound was improved to 73% (45/62) but specificity fell to 30% (23/77). The entire histologic composition of a section was correctly diagnosed by ultrasound in only 42% of the 139 sections. Further technical improvements are therefore mandatory before intracoronary ultrasound will be able to provide a reliable analysis of plaque composition, especially of the lipid content.