Computer-Based Automated Left Atrium Segmentation and Volumetry from ECG-Gated Coronary CT Angiography Data: Comparison with Manual Slice Segmentation and Ultrasound Planimetric Methods

Abstract

Enlargement of the left atrium is a risk factor for cardiovascular or cerebrovascular events. We evaluated the performance of prototype software for fully automated segmentation and volumetry of the left atrium. In 34 retrospectively ECG-gated coronary CT angiography scans, the end-systolic (LAVsys) and end-diastolic (LAVdia) volume of the left atrium was calculated fully automatically by prototype software. Manual slice segmentation by two independent experienced radiologists served as the reference standard. Furthermore, two independent observers calculated the LAV utilizing two ultrasound planimetric methods ("area length" and "prolate ellipse") on CTA images. Measurement periods were compared for all methods. The left atrial volumes calculated with the prototype software were in excellent agreement with the results from manual slice segmentation (r = 0.97 - 0.99; p < 0.001; Bland-Altman) with excellent interobserver agreement between both radiologists (r = 0.99; p < 0.001). Ultrasound planimetric methods clearly showed a higher variation (r = 0.72 - 0.86) with moderate interobserver agreement (r = 0.51 - 0.79). The measurement period was significantly lower with the software (267 ± 28 sec; p < 0.001) than with ultrasound methods (431 ± 68 sec) or manual slice segmentation (567 ± 91 sec). The prototype software showed excellent agreement with manual slice segmentation with the least time consumption. This will facilitate the routine assessment of the LA volume from coronary CTA data and therefore risk stratification.