Purpose: To investigate the accuracy of x-ray grating interferometry phase-contrast (PC) imaging for the characterization of human coronary artery plaque.

Materials and methods: PC and conventional absorption computed tomographic (CT) imaging was performed ex vivo in this institutional review board-approved study in 40 human coronary artery segments by using a synchrotron radiation source. Qualitative analyses and calculations of image quality (McNemar test), plaque components (McNemar test), and plaque classification (Cohen κ test) according to the American Heart Association classification were performed in 38 plaques detected at histopathologic examination (reference standard). Quantitative measurements of plaque components (ie, collagen, lipids, smooth muscle, and calcifications) were compared among PC and absorption images by using analysis of variance for repeated measures with post hoc Bonferroni correction.

Results: Image quality was superior in PC (median image score, 1) in all cases (100%) compared with absorption imaging (median image score, 3) (P < .001). Plaque components were detected by means of PC without significant differences (seven of seven calcifications, 22 of 22 plaques with collagen and smooth muscle cells, P > .99; 29 of 29 plaques with lipids, P = .10) with histopathologic findings, whereas absorption imaging was used to detect calcifications (seven of seven, P > .99) without statistical differences only (nine of 29 plaques with lipids, 0 of 22 plaques with collagen and smooth muscle cells, P < .001). Accuracy for plaque stage assessment with PC (early vs advanced) was 100%, and characterization was correct in 33 of 38 plaques (87%), while conventional absorption imaging allowed correct characterization of seven plaques only (18%, P < .001). PC CT numbers were significantly different (P < .05) for all plaque components (mean for calcifications, 1236 HU ± 69; collagen, 78 HU ± 24; lipids, -18 HU ± 23; and smooth muscle cells, 34 HU ± 12), whereas absorption images showed significant differences (P < .001) between calcifications (1336 HU ± 241) and other plaque components, but not for collagen (22 HU ± 13), lipids (-15 HU ± 14), and smooth muscle (13 HU ± 9) (P > .99).

Conclusion: PC imaging allows accurate characterization of human coronary artery plaques and quantitative assessment of plaque components, thereby outperforming absorption imaging.