Move Over There: One-Click Deformation Correction for Image Fusion During Endovascular Aortic Repair

Abstract

Fusing intraoperative X-ray with information from preoperative computed tomography for endovascular aortic repair has been shown to reduce radiation exposure, need for contrast agent, and procedure time. However, due to the instruments inserted during the intervention, the vasculature deforms and the fusion loses accuracy. In this paper, we propose an approach to use minimal user input obtained from a single 2D image for deformation correction in 3D. We integrate the 2D positional information as a projective constraint in an as-rigid-as-possible deformation model that allows for intraoperative correction of the deformation. Our method achieves clinically relevant accuracies while keeping user inputs to a minimum. We are able to recover the deformation at the right and left internal iliac bifurcation up to a 3D error of 1.9mm, with an error of 0.5mm orthogonal to the viewing direction, and an error of 1.7mm in depth, while keeping the mean computation time below 6s.