Encrypted login | home

Program Information

Use of a Deformable Ultrasound Phantom for Tracking Algorithm Development and Validation

no image available
A Shepard

A Shepard*, C Matrosic , J Zagzebski , B Bednarz , University of Wisconsin, Madison, WI

Presentations

SU-E-601-5 (Sunday, July 30, 2017) 1:00 PM - 1:55 PM Room: 601


Purpose: To develop an ultrasound phantom and testbed with the ability to produce deformable feature motion for validating tracking algorithms that account for the presence of non-rigid motion.

Methods: A 7.5-cm diameter, cylindrical, gelatin-based phantom was produced that incorporated a single hyperechoic inclusion spanning the diameter of the phantom. The phantom was constructed to allow for deformation along the axial dimension while restricting motion along the radial dimension. To create a fully automated testbed to produce deformation sequences, the gel phantom was coupled with a 1D motion stage and imaged using a GE Vivid E9 ultrasound scanner (4V transducer). The transducer and ultrasound phantom were fixed within the setup, with the transducer coupled to the top edge of the phantom. The 1D motion stage was used to move a sphere into the bottom edge of the phantom, thus creating controlled deformations in the feature. The hyperechoic inclusion consisted of a water and corn starch solution to enhance contrast between the feature and the surrounding gel. The acquired motion sequence was retrospectively analyzed using a block-matching based contour tracking algorithm, and the performance was analyzed relative to ground truth segmentations performed on 5% of the acquired frames.

Results: The contour tracking performed by the algorithm demonstrated a mean Dice coefficient of 0.93 (0.89-0.96) when compared to manual segmentations. The lowest Dice coefficients were observed during times at which the magnitude of the deformation was largest, indicating the potential for algorithm improvement in the presence of large deformations.

Conclusion: The pairing of the phantom and testbed displayed feature deformation that can aid in tracking algorithm development by producing more realistic, non-rigid feature motion. Further development of the inclusion shape and the ultrasonic properties of the gel will lead to a better representation of features seen on ultrasound B-mode images of the body.

Funding Support, Disclosures, and Conflict of Interest: This work was supported by NIH grant R01CA190298.


Contact Email: