Pose-aware C-Arm Calibration and Image Distortion Correction for Guidewire Tracking and Image Reconstruction

Abstract

Image intensifiers, also known as C-arms, are important and low-cost tools for surgeons to guide minimally-invasive procedures. However, image intensifiers suffer from several distortions that can impede their ability to provide accurate guidance. These distortions can be misleading during an automated minimally-invasive surgery where the accurate shape-estimation of the robot is essential. Since the distortion strongly depends on the orientation of the C-arm during image acquisition, we propose an approach for distortion correction in combination with a calibration procedure to precisely estimate its orientation. To estimate an accurate distortion correction function, we take images of a calibration grid in 258 different C-arm orientations and apply polynomial regression on these images. The C-arm is calibrated using an external camera along with optical flow and point-correspondence-based matching to allow sufficient pose estimation. Our C-arm tracking algorithm estimates the pose of the C-arm with a mean absolute deviation of 0.2° and 0.3° for 5° and 10° relative motion. The proposed C-arm calibration procedure allows the positioning of the C-arm within an error range of 0.5°. The resulting distortion correction leads to a mean pixel displacement of 0.30 pixel.