Premium
SU‐D‐207‐06: Clinical Validations of Shading Correction for Cone‐Beam CT Using Planning CT as a Prior
Author(s) -
Tsui T,
Wei J,
Zhu L
Publication year - 2015
Publication title -
medical physics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.473
H-Index - 180
eISSN - 2473-4209
pISSN - 0094-2405
DOI - 10.1118/1.4923907
Subject(s) - truebeam , radiation treatment planning , cone beam computed tomography , image guided radiation therapy , cone beam ct , robustness (evolution) , computer science , medical imaging , nuclear medicine , ground truth , radiation therapy , artificial intelligence , medicine , computer vision , computed tomography , radiology , beam (structure) , optics , physics , linear particle accelerator , biochemistry , chemistry , gene
Purpose: Current cone‐beam CT (CBCT) images contain severe shading artifacts mainly due to scatter, hindering their quantitative use in current radiation therapy. We have previously proposed an effective shading correction method for CBCT using planning CT (pCT) as prior knowledge. In this work, we investigate the method robustness via statistical analyses on studies of a large patient group and compare the performance with that of a state‐of‐the‐art method implemented on the current commercial radiation therapy machine ‐‐ the Varian Truebeam system. Methods: Since radiotherapy patients routinely undergo multiple‐detector CT (MDCT) scans in the planning procedure, we use the high‐quality pCT as “free” prior knowledge for CBCT image improvement. The CBCT image with no correction is first spatially registered with the pCT. Primary CBCT projections are estimated via forward projections of the registered image. The low frequency errors in the projections, which stem from mainly scatter, are estimated by filtering the difference between original line integral and the estimated scatter projections. The corrected CBCT image is then reconstructed from the scatter corrected projections. The proposed method is evaluated on 40 cancer patients. Results: On all patient images, we compare errors on CT number, spatial non‐uniformity (SNU) and image contrast, using pCT as the ground truth. T‐tests show that our algorithm improves over the Varian method on CBCT accuracies of CT number and SNU with 90% confident. The average CT number error is reduced from 54.8 HU on the Varian method to 40.9 HU, and the SNU error is reduced from 7.7% to 3.8%. There is no obvious improvement on image contrast. Conclusion: Large‐group patient studies show that the proposed pCT‐based algorithm outperforms the Varian method of the Truebeam system on CBCT shading correction, by providing CBCT images with higher CT number accuracy and greater image uniformity.