Calibration of Freehand Ultrasound Based on YOLOv8x Keypoints Detection Network
校准
计算机科学
人工智能
计算机视觉
数学
统计
作者
Fan Peng,Shangqi Cui,Binbin Wang,D. Q. Wang,Zhe Shi,Lanlin Li,Zhi Yang
标识
DOI:10.1145/3644116.3644147
摘要
Ultrasound 3D reconstruction could be used for realizing non-ionizing radioactive intraoperative ultrasound guidance, in which spatial calibration of the ultrasound probe was a critical step. In this study, a freehand ultrasound approach was used and an N-wire phantom was used for spatial calibration. Instead of manually picking the bright fiducial points in the images as the previous studies reported in the literatures, this work identified fiducial points automatically based on the YOLOv8x keypoints detection, which improved the precision and efficiency of probe spatial calibration over the manual methods. The experimental results showed that the calibration accuracy was 2.02mm, and the processing time on each frame was 0.075s.