Variability and Accuracy of Coronary CT Angiography Including Use of Iterative Reconstruction Algorithms for Plaque Burden Assessment As Compared with Intravascular Ultrasound-an Ex Vivo Study

Objectives: To systematically assess inter-technique and inter-/intra-reader variability of coronary CT angiography (CTA) to measure plaque burden compared with intravascular ultrasound (IVUS) and to determine whether iterative reconstruction algorithms affect variability.

Methods: IVUS and CTA data were acquired from nine human coronary arteries ex vivo. CT images were reconstructed using filtered back projection (FBPR) and iterative reconstruction algorithms: adaptive-statistical (ASIR) and model-based (MBIR). After co-registration of 284 cross-sections between IVUS and CTA, two readers manually delineated the cross-sectional plaque area in all images presented in random order.

Results: Average plaque burden by IVUS was 63.7 ± 10.7% and correlated significantly with all CTA measurements (r = 0.45-0.52; P < 0.001), while CTA overestimated the burden by 10 ± 10%. There were no significant differences among FBPR, ASIR and MBIR (P > 0.05). Increased overestimation was associated with smaller plaques, eccentricity and calcification (P < 0.001). Reproducibility of plaque burden by CTA and IVUS datasets was excellent with a low mean intra-/inter-reader variability of <1/<4% for CTA and <0.5/<1% for IVUS respectively (P < 0.05) with no significant difference between CT reconstruction algorithms (P > 0.05).

Conclusion: In ex vivo coronary arteries, plaque burden by coronary CTA had extremely low inter-/intra-reader variability and correlated significantly with IVUS measurements. Accuracy as well as reader reliability were independent of CT image reconstruction algorithm.

Key Points: • IVUS is deemed the gold standard in-vivo coronary plaque assessment • But coronary CT angiography findings correlate strongly with IVUS results • Coronary CT angiography now allows plaque quantification close to IVUS • Iterative image reconstruction algorithms do not alter accuracy or reproducibility • Plaque quantification is more challenging in smaller eccentric calcified lesions.