Automated detection of aortic root landmarks in preprocedure CT angiography images for transcatheter aortic valve implantation patients

Transcatheter aortic valve implantation provides a minimal invasive treatment in patients with severe aortic stenosis. CT Angiography is used for the pre-operative planning, in which the accessibility of the aorta-femoral tract for the catheter and the prosthetic type and size can be determined. Preprocedure planning includes the determination of annulus radius, area and coronary ostia to annulus distance. These measurements use the location of five landmarks; the two coronary ostia and the three hinge points. Automatic landmarks detection is beneficial to speed up the calculation of the sizing parameters. In this paper, we introduce an automated approach to extract the aortic root landmarks and calculate sizing parameters. Our proposed algorithm has a high accuracy in comparison with the manual reference with a mean point-to-point error of 2.47 mm in 20 patients; where the interobserver variation had a mean point to point of 2.30 mm. With the high accuracy shown, the proposed method can be introduced in clinical practice.