Automated correlation of intravascular ultrasound images with angiography: an alternative to three-dimensional reconstructions

Abstract

Contrast angiography is the present standard for imaging of the vascular system, but intravascular ultrasound (IVUS) provides advantages for the study of arterial wall pathology. To enhance such studies, we developed a computer algorithm to automatically correlate IVUS images with their corresponding locations on contrast angiograms, and to display both images in a side-by-side format. We verified the accuracy of this computer algorithm using vascular models of the aorta and ileo-femoral vessels. Artificial plaques were placed at various sites throughout the models. The models were filled with iodinated contrast and radiographic images were obtained. Serial timed pullbacks using a 20 @ ultrasound catheter were performed in each model and acquired images were sampled, digitized, and stored. A computer algorithm written using a package for image processing and mathematical analysis accessed the stored images for processing. At least one IVUS image containing a known landmark was then visually matched with its position on its corresponding angiogram. The computer algorithm automatically correlated the remaining IVUS frames with their expected locations on the angiogram. The IVUS images could then be scrolled through while their corresponding position on the angiogram was displayed. Error was measured after using the procedure with various numbers of correlation points. There was an inverse relation between the number of IVUS points visually matched in the models and the measured errors. The maximum error was 4 mm over a total pullback distance of 130 mm (relative error 3%) with only one correlation point. This declined to less than 3 mm when three or more points were used. These results demonstrate that computer based correlation of IVUS images with angiography provides an accurate alternative to three-dimensional reconstruction algorithms.