International Journal of Radiation Oncology*Biology*Physics
ePoster AbstractVesselness-Based Deformable Registration Algorithms Can Reduce Landmark Errors in the Registration of Pulmonary Computed Tomography Images Before and After Radiation Therapy
Section snippets
Purpose/Objective(s)
Deformable registration algorithms can aid physicians in evaluating radiation therapy (RT) response after radiation therapy. However, registering pre-RT and post-RT images is challenging due to the presence of post-RT fibrosis. Large fibrotic regions may increase landmark registration errors and cause pronounced anatomic distortions around the tumor using traditional intensity-based deformable registration algorithms. We hypothesize that vesselness-based deformable registration algorithms,
Materials/Methods
A database of twenty-two consecutive patients, who had two courses of RT with available helical computed tomography scans, was assembled. Image registrations were performed between pre-RT and post-RT scans with two deformable registration algorithms: a control intensity-based deformable registration algorithm, and a vesselness-based deformable registration algorithm. Both algorithms were implemented on the Elastix platform. Registrations were evaluated using the landmark error metric, which is
Results
The median landmark errors for the control and vesselness algorithms were 1.9 (interquartile range [IQR] 1.7-2.6) and 1.7 (1.4-2.3), respectively, and their difference was statistically significant (P = 0.03). Respective TDI values (lower is better) were 0.72 (0.36-1.01) and 0.25 (0.21-0.31). Their difference was also statistically significant (P < 0.01). Visual inspection of images showed improved morphological preservation of pre-RT tumor with post-RT fibrotic regions, as well as
Conclusion
Vesselness-based deformable registration algorithm may play a promising role in the registration of images before and after RT. This algorithm may pave the way for enhanced pulmonary function assessment on a per-voxel basis after RT.
References (0)
Cited by (0)
Author Disclosure: M. King: None. G. Xiong: None. P.G. Maxim: Stock Options; TibaRay, Inc. M. Diehn: Research Grant; Varian Medical Systems. B.W. Loo: Research Grant; Varian Medical Systems, RaySearch Laboratories. Stock Options; TibaRay, Inc. Practice Parameters; American College of Radiology. Small cell lung cancer; National Comprehensive Cancer Network. L. Xing: Research Grant; Varian Medical Systems.