A novel method for an automatic 3D reconstruction of coronary arteries from angiographic images

This paper presents a new application of the shape from silhouette (SFS) method for the 3D reconstruction of coronary arteries with Kawasaki disease from angiographic images. The silhouettes of the arteries were first segmented from the angiographic images using the Frangi filter followed by thresholding then a morphologic erosion step. Finally, a volume-based reconstruction approach was performed to infer the visual hull representing the 3D volume of the segmented arteries. The proposed method was first validated on simulated data to determine the optimal number of views that provide a 3D reconstruction with an adequate precision. The sensitivity of the proposed method to calibration errors was also evaluated. Furthermore, a clinical evaluation on four patients with Kawasaki disease showed that using three views in addition to the standard AP and LAT views does not improve the accuracy of the 3D reconstruction. Therefore, this study showed that an online 3D reconstruction of the coronary arteries, requiring less than one minute, during the acquisition could help to rationalize the number of views required for the 3D assessment of aneurysms.