An iterative closest point framework for ultrasound calibration

Abstract

We introduce an Iterative Closest Point framework for ultrasound calibration based on a hollow-line phantom. The main novelty of our approach is the application of a hollow-tube fiducial made from hyperechoic material, which allows for highly accurate fiducial localization via both manual and automatic segmentation. By reducing fiducial localization error, this framework is able to achieve sub-millimeter target registration error. The calibration phantom introduced can be manufactured inexpensively and precisely. Using aMonte Carlo approach, our calibration framework achieved 0.5mm mean target registration error, with a standard deviation of 0.24 mm, using 12 or more tracked ultrasound images. This suggests that our framework is approaching the accuracy limit imposed by the tracking device used.