Program Information
Development of a Programmable Motion Testbed for the Validation of Ultrasound Tracking Algorithms
A Shepard*, C Matrosic , J Zagzebski , B Bednarz , University of Wisconsin, Madison, WI
Presentations
SU-G-JeP1-7 (Sunday, July 31, 2016) 4:00 PM - 4:30 PM Room: ePoster Theater
Purpose: To develop an advanced testbed that combines a 3D motion stage and ultrasound phantom to optimize and validate 2D and 3D tracking algorithms for real-time motion management during radiation therapy.
Methods: A Siemens S2000 Ultrasound scanner utilizing a 9L4 transducer was coupled with the Washington University 4D Phantom to simulate patient motion. The transducer was securely fastened to the 3D stage and positioned to image three cylinders of varying contrast in a Gammex 404GS LE phantom. The transducer was placed within a water bath above the phantom in order to maintain sufficient coupling for the entire range of simulated motion. A programmed motion sequence was used to move the transducer during image acquisition and a cine video was acquired for one minute to allow for long sequence tracking. Images were analyzed using a normalized cross-correlation block matching tracking algorithm and compared to the known motion of the transducer relative to the phantom.
Results: The setup produced stable ultrasound motion traces consistent with those programmed into the 3D motion stage. The acquired ultrasound images showed minimal artifacts and an image quality that was more than suitable for tracking algorithm verification. Comparisons of a block matching tracking algorithm with the known motion trace for the three features resulted in an average tracking error of 0.59 mm.
Conclusion: The high accuracy and programmability of the 4D phantom allows for the acquisition of ultrasound motion sequences that are highly customizable; allowing for focused analysis of some common pitfalls of tracking algorithms such as partial feature occlusion or feature disappearance, among others. The design can easily be modified to adapt to any probe such that the process can be extended to 3D acquisition. Further development of an anatomy specific phantom better resembling true anatomical landmarks could lead to an even more robust validation.
Funding Support, Disclosures, and Conflict of Interest: This work is partially funded by NIH grant R01CA190298
Contact Email: