SU‐F‐J‐78: Characterization of Deformable Image Registration for the Pelvic Region Based On Prostate Shifting, Image Noise and the Existence of Implanted Fiducial Markers

摘要

Purpose: To assess the impact of prostate shifting, image noise and the existence of fiducial markers on deformable image registration (DIR) performance in the pelvic region using original digital phantoms. Methods: A digital phantom of the pelvic region of a prostate cancer patient was created using contoured anatomy as a base image set. Two simulated fiducial markers were implanted at the apex and base of the prostate. With and without markers, the prostate was shifted by 6.00 and 10.0 mm superior, 5.63 and 9.38 mm posterior, and 5.63 and 9.38 mm left, respectively. We also added image noise with a standard deviation (SD) of 9, 21, or 28. A phantom with a noise SD of 9 and zero prostate shift was defined as the target, and the other phantoms were defined as the sources. DIR was performed using MIM Maestro (MIM Software Inc., Cleveland, OH, USA) on the digital phantoms. A dice similarity coefficient (DSC) was used as the index of correspondence level of the prostates on the target images and the results of the DIR. Also, we calculated the motion distance of the fiducial markers. Results: The DSC decreased with increasing prostate distance. The degree of correspondence of the fiducial markers and the DIR performance were worse when the bladder volume changed after the prostate shifted in the superior direction than when it shifted in the other directions. Changes in the image noise and the existence of fiducial markers produced minimal impact on the DIR results. Conclusion: We found that DIR was not affected by image noise, and the reproduction of prostate position by maintaining the bladder volume was important to preserve DIR performance. Further, the existence of two fiducial markers had no impact on DIR performance. This work was supported by Japan Society for the Promotion of Science (JSPS) KAKEN Grant‐in‐Aid for Scientific Research (C) Grant number 26460724.