Abstract 15317: Automatic Quantification and Characterization of Coronary Plaque with Computed Tomography Coronary Angiography: A Comparison with Intravascular Ultrasound Virtual Histology

Abstract

Purpose Several studies have identified plaque characteristics on CTA associated with plaque vulnerability. At present only visual assessment of plaque characteristics is possible. An accurate quantitative approach on CTA plaque vulnerability would improve reproducibility and allows for better risk stratification. However, at present such an automatic CTA plaque characterization tool is unavailable. Therefore, the aim of the present study was to create an automatic CTA tissue characterization tool and to validate this tool against IVUS VH. Methods A total of 44 Patients underwent CTA prior to IVUS VH. First, automated lumen and vessel wall contour detection was performed in both modalities. Then, quantitative CTA (QCT) was performed. Per lesion stenosis parameters and plaque volumes were assessed. Using predefined HU thresholds, CTA plaque volume was differentiated in 4 different plaque types (necrotic core (NC), dense calcium (DC), fibrotic (FI) and fibro-fatty tissue (FF). At the identical level of the coronary, using a dedicated algorithm to fuse the CTA and IVUS VH images, the same parameters were derived from IVUS VH. Bland-Altman analysis was performed to assess the bias and 95% limits of agreement between QCT and IVUS VH. Results Assessment of plaque volume using QCT in 108 lesions showed good correlation with IVUS VH (r =0.900, P<0.001) (figure 1). For the plaque volume bias was 36 mm3 with limits of agreement from -48 to 121 mm3. The agreement in both FF and FI volume on IVUS VH and QCT was (r =0.704, P<0.001) and (r =0.787, P<0.001) respectively with corresponding bias and 95% limits of agreement of 20 (-15; 56) and 20 (-36; 77). Furthermore, the correlation between NC and DC in both modalities was (r =0.479, P<0.001) and (r =0.733, P<0.001) with the following bias and 95% limits of agreement of 1 (-26; 28) and 8 (-32; 49), respectively. Conclusion Automatic CTA tissue characterization is feasible using a dedicated software tool.