Improving detection of coronary morphological features from digital angiograms. Effect of stenosis-stabilized display

We have developed a digital display method that stabilizes the motion of a stenosis in sequential frames of a coronary angiogram, allowing it to be scrutinized at high display frame rates. The purpose of this study was to determine whether this technique improves visual detection of low-contrast luminal morphological features.

An observer detection study was conducted using computer-simulated arterial segments containing known target features, inserted into clinical digital coronary angiograms. Four observers performed a forced-choice detection of a simulated filling defect in each of 320 angiograms using the conventional and stenosis-stabilized dynamic displays (at 7.5, 15, and 32 frames per second) and a single-frame static display (total of 8960 detections). In a second simulated clinical task, three observers detected a bridging stenotic lumen in 600 angiograms using the two displays (3600 detections). In a third experiment, two angiographers rated the likelihood of intraluminal thrombus in 89 right coronary digital angiograms by consensus reading with both dynamic displays. Detectability of the simulated filling defect was similar for both dynamic display methods at 7.5 frames per second (averaging twice that for static images). As display rate was increased to 32 frames per second, detectability for the conventional display declined, whereas the stabilized display detectability increased for all observers (P < .05). On average, stabilization allowed detection of filling defects equivalent to a 71% increase in feature contrast. Response time for the conventional display averaged 12.9 +/- 4.7 seconds. For the stenosis-stabilized display, response time fell with increased frame rate (P < .05) to 4.9 +/- 1.2 seconds at 32 Hz, similar to the time for static images (4.6 +/- 0.8 seconds). The detectability of the bridging stenotic lumen was increased by 62% with the stabilization compared with conventional dynamic display (P < .00001). Consensus reading of coronary angiograms showed differences between the two dynamic display methods (kappa = 0.11) that may be explained by an improvement in observer uncertainty. A rating of definite for thrombus present or absent was more frequent with the stabilized display (39% versus 15%, P < .0001).

These data suggest that stabilized display of coronary angiograms significantly increases detectability, reduces the time required for detection, and improves observer uncertainty for the presence of small luminal morphological features. The method of angiographic display may thus have an impact on clinical coronary angiographic interpretation.