MENU

Click here to

×

Are you sure ?

Yes, do it No, cancel

An Easy to Create In-House Phantom for TG-132 Deformable Image Registration Evaluation

T Martin*, D Saenz , P Myers , K Rasmussen , S Stathakis , N Papanikolaou , N Kirby , University of Texas HSC SA, San Antonio, TX

Presentations

(Sunday, 7/14/2019)  

Room: ePoster Forums

Purpose: Digital phantom tests are one of the tools recommended by TG-132 for commissioning a deformable image registration (DIR) software in radiotherapy. The purpose of this study was to implement an easy to replicate in-house digital phantom test without the use of advanced image processing or digital phantom software packages.

Methods: Two patient CT image sets, of empty and full bladder, were used for creating the digital phantoms. One was created in MIM software and the other in Velocity Medical Solutions software. The ground-truth in this study was established by deforming the moving image (empty bladder) to the stationary image (full bladder) using a deformation algorithm in each software, and then resampling it. The applied DVF is the ground-truth between the moving and resampled image. A clinical DIR algorithm could then be applied between these images. MIM only has one DIR algorithm so it was used for both test and clinical algorithm, while in Velocity two separate DIR algorithms were used. The DVF for the ground truth and the clinical DIR were exported and then the percentage of voxels with less than a 2 mm error and the maximum error were compared. The TG-132 recommended tolerances for these are 95% and 5 mm, respectively.

Results: The percentage of voxels with less than a 2 mm error for MIM was 98% with a maximum error of 14.5 mm and for Velocity it was 64% with a maximum error of 11.5 mm.

Conclusion: This study demonstrated an easy to create digital phantom for evaluation and commissioning of DIR algorithms. MIM overall produced more accurate results, but this could be caused by utilizing the same DIR algorithm. Both MIM and Velocity had larger than the TG recommended tolerances. However, these numbers could still be useful for setting baseline performances.

Keywords

Deformation, Digital Imaging

Taxonomy

Not Applicable / None Entered.

Contact Email