BEST IN PHYSICS (JOINT IMAGING-THERAPY) - Evaluation of Deformation Algorithm Accuracy with a Two-Dimensional Anatomical Pelvic Phantom
N Kirby*, C Chuang, J Pouliot, UC San Francisco, San Francisco, CAWE-E-213CD-1 Wednesday 2:00:00 PM - 3:50:00 PM Room: 213CD
Purpose: To objectively evaluate the accuracy of 11 different deformable registration techniques for bladder filling.
Methods: The phantom represents an axial plane of the pelvic anatomy. Urethane plastic serves as the bony anatomy and urethane rubber with three levels of Hounsfield units (HU) is used to represent fat and organs, including the prostate. A plastic insert is placed into the phantom to simulate bladder filling. Nonradiopaque markers reside on the phantom surface. Optical camera images of these markers are used to measure the positions and determine the deformation from the bladder insert. Eleven different deformable registration techniques are applied to the full- and empty-bladder computed tomography images of the phantom to calculate the deformation. The applied algorithms include those from MIMVista Software and Velocity Medical Solutions and 9 different implementations from the Deformable Image Registration and Adaptive Radiotherapy Toolbox for Matlab. The distance to agreement between the measured and calculated deformations is used to evaluate algorithm error. Deformable registration warps one image to make it similar to another. The root-mean-square (RMS) difference between the HUs at the marker locations on the empty-bladder phantom and those at the calculated marker locations on the full-bladder phantom is used as a metric for image similarity.
Results: The percentage of the markers with an error larger than 3 mm ranges from 3.1% to 28.2% with the different registration techniques. This range is 1.1% to 3.7% for a 7 mm error. The least accurate algorithm at 3 mm is also the most accurate at 7 mm. Also, the least accurate algorithm at 7 mm produces the lowest RMS difference.
Conclusions: Different deformation algorithms generate very different results and the outcome of any one algorithm can be misleading. Thus, these algorithms require quality assurance. The two-dimensional phantom is an objective tool for this purpose.