Abstract Background Anomalous aortic origin of a coronary artery (AAOCA) is associated with increased risk of myocardial ischemia and sudden cardiac death. The intramural (IM) segment is a key factor given risk of dynamic compression and adverse cardiac events upon exercise. Existing computed tomography angiography (CTA)-based methods show variable agreement with surgical findings, an example of which is shown (Fig. A). No studies to date have utilized computational modeling for assessing intramural length. Purpose To develop and validate a computational technique for automatically estimating IM length from CTA that is relatively accurate and reliable compared to surgical measurements, improving precision and consistency in risk assessment. Methods A single-institution retrospective study of CTA in 58 surgical AAOCA patients was performed. De-identified CTA data were used to create meshes of the aorta (Fig. B). Next, the centerline of the coronary artery was generated (Fig. B). IM length was estimated using a signed distance function from the vessel centerline to the aortic lumen to generate a transition point where the coronary artery exits the aortic wall (Fig. C). Accuracy was assessed using the root-mean-square error (RMSE), calculated against the surgically measured IM length. The RMSE of the computational model was compared to the RMSE of radiologic estimates and stratified by left (L)- and right (R)-AAOCA (Fig. D, E). Box-and-whisker plots were used to compare the errors between the computational and radiologic methods, as measured against the surgical measurements (Fig. F). Results This study included 49 R-AAOCA and 9 L-AAOCA surgical patients. The overall RMSE for the computational approach was 3.44 mm, compared to 3.18 mm for the radiologic approach. For L-AAOCA, the RMSE was 3.61 mm by computational compared to 4.7 mm by radiologic estimates. For R-AAOCA, the RMSE was 3.41 mm by computational compared to 2.82 mm by radiologic method. Conclusion The computational approach produced RMSE values comparable to the radiologic method. In addition, our technique was more accurate in L-AAOCA cases. This novel, automatic, and noninvasive approach provides an accurate and reliable estimation of IM length. Further validation in larger cohorts and integration into clinical workflows are needed and could improve AAOCA risk stratification and patient management.
Ferrino et al. (Sat,) studied this question.