An iterative closest point framework for ultrasound calibration
Document Type
Conference Proceeding
Publication Date
1-1-2015
Journal
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume
9365
First Page
69
Last Page
79
URL with Digital Object Identifier
10.1007/978-3-319-24601-7_8
Abstract
© Springer International Publishing Switzerland 2015. 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.