유방 병변 제거술을 위한 비마커 기반 AR 수술 프레임워크

초록

In this paper, we propose a markerless AR surgical framework for breast lesion removal using depth sensor and 3D breast CT. In order to acquire a mesh to be used for registration, a patient mesh on the real coordinate system is acquired through a patient 3D scan using a depth sensor. The patient mesh on the virtual coordinate system is obtained by performing contrast-based skin segmentation in 3D breast CT and generating a 3D mesh. Based on the gradient in the segmented skin area, detect the nipple, set the ROI based on this, and select the vertices on the virtual coordinate system to be used for matching. The mesh on the real and virtual coordinate systems align the primary positions through the matching of the center of mass, and applies the ICP method for registration. As a result of experiments using data from 10 patients, the skin segmentation accuracy was DSC 98.34%, the nipple detection error was 2.99 mm, and the registration error was 5.01 mm. In addition, the registration error using 10 phantom data was 3.42 mm, showing high accuracy of AR visualization.

키워드