Quantitative Assessment of Calibration Motion Profiles in Robotic-assisted Ultrasound System

Abstract

3D ultrasound(US) reconstruction has grown quickly in preoperative disease diagnosis, intraoperative surgical navigation and postoperative treatment assessment. US image calibration is a crucial step to produce accurate 3D representations from 2D images. Several image calibration methods have been developed for freehand and robotic systems. During calibration, the scanning motions can vary a lot and typically have a large impact on the resulting calibration and subsequent reconstruction quality. However, the adopted calibration motions are not detailed extensively in previous studies. This complicates reproduction of the obtained results. Moreover, the influence of the different scanning motions on the calibration accuracy is hardly investigated. Therefore, it is difficult to devise optimal scanning motion profiles for automatic robotic ultrasound calibration. This paper studies calibration and reconstruction results with different motion profiles employed in robotic US image calibration. Then, the calibration procedure was performed with a sphere phantom. The performance of the calibration was validated by assessing the reconstruction quality on a 3D printed mock-up model with two quantitative measurements for the geometric representation error and the 3D localization error. The geometric representation error of the reconstruction is within 1 mm by using the different combination of motion profiles. However, the 3D localization error changes with the various motion profiles. By using the proposed motion profile, the RMSE of the reconstructed model could be reduced to 2.18 mm.

Type
Publication
2022 International Symposium on Medical Robotics (ISMR)