Session S94.1

Volumetric Quantification of Myocardial Perfusion Using Analysis of Multi-Detector Computed Tomography 3D Datasets

N Kachenoura*, F Veronesi, JA Lodato, C Corsi, R Mehta,
B Newby, RM Lang, V Mor-Avi

University of Chicago
Chicago, IL, USA

Although the ability of multi-detector computed tomography (MDCT) to detect perfusion defects associated with myocardial infarction has been demonstrated, this methodology is based on visual interpretation of selected 2D slices, which requires adjustment of contrast windows, and is thus subjective and sometimes difficult. Accordingly, we sought to develop a new technique for volumetric quantification of myocardial perfusion from 3D datasets and test its ability to accurately determine the presence, extent and severity of perfusion defects using resting nuclear myocardial perfusion imaging (MPI) as a reference. Methods. We studied 44 patients undergoing MDCT coronary angiography (CTCA): control group (N=15) and study group (N=29). MDCT data acquired for CTCA were analyzed using custom software. First, left ventricular (LV) myocardium was identified using level-set technique and 16 standardized 3D myocardial segments were automatically defined. For each short axis slice of the 3D dataset, myocardial x-ray attenuation calculated along radial profiles and divided by LV cavity attenuation to normalize for differences between patients were used to generate bull’s eye display of myocardial perfusion. Additionally, distribution of x-ray attenuation in individual segments was used to calculate a segmental quantitative index of extent and severity of perfusion abnormality, QH. For both analyses, values obtained in the control group were used to compensate for segmental heterogeneity in x-ray attenuation. Visual interpretation of MDCT bull’s eyes was compared with rest MPI scores on a coronary territory and patient basis using kappa statistics. For each vascular territory and each patient, segmental QH indices were summed and correlated with rest MPI summed scores. Additionally, QH was used for objective detection of perfusion defects. Results. Visual analysis of MDCT-derived bull’s eyes accurately detected perfusion defects in agreement with MPI (kappa=0.70 by territory; 0.79 by patient). Quantitative data were in good agreement with MPI, as reflected by: (1) correlation of 0.87 (territory) and 0.84 (patient) between summed QH and MPI scores, (2) area under ROC curve 0.87 with sensitivity of 0.79-0.92, specificity 0.83-0.91, and accuracy 0.83-0.89 for objective detection of abnormalities. Conclusions. Our new technique for volumetric analysis of MDCT images allows accurate and objective detection of perfusion defects. This perfusion information can be obtained without additional radiation or contrast load, and may aid in elucidating the significance of coronary lesions.

(Abstract Control Number: 223)