Quantitative Cone-Beam CT Imaging in Radiation Therapy Using Planning CT as a Prior: Clinical Assessments
HY Zhang1*, L Liu1, H Wu2, HZ Yue2, TY Niu3, L Zhu3, (1) Anhui Province Hospital, Heifei, China, (2) Beijing Cancer Hospital & Institute, Beijing, China, (3) Georgia Institute of Technology, Atlanta, GAWE-C-WAB-7 Wednesday 10:30AM - 12:30PM Room: Wabash Ballroom
Cone-beam CT (CBCT) imaging is on increasing demand for advanced image guided radiation therapy (IGRT). However, scatter contamination causes large CT number errors. To improve CBCT imaging for quantitative use, we design a planning CT (pCT) based software package readily transferrable to the current CBCT system of radiation therapy machine with high-quality CT images and short computation time. The clinical benefits using the proposed method are assessed on the pelvis patients in terms of the accuracy of CT number and target volume delineation.
Planning CT (pCT) image is routinely acquired on a patient for planning purposes in radiation therapy. Our software estimates the primary signals of CBCT projections via forward projection on the registered pCT image, and then obtains the low-frequency errors in CBCT raw projections by subtracting the estimated primary signals and low-pass filtering. To facilitate clinical applications, the computation is accelerated on a NVidia Tesla C2075 GPU card. The performance of the software is evaluated on three pelvis patients with CBCT projections acquired from the on-board imager (OBI) installed on Varian Linac systems. We also analyze the inter-observer contouring accuracy of target volumes on the CBCT images of one patient after the correction.
After GPU acceleration, the execution time of the software is reduced from over 10 hours to 25 minutes. Compared to the TrueBeam reconstruction, our software reduces the maximum CT number error from over 70 HU to below 15 HU. We also reduce the delineation error from 30% to below 8%.
We design a clinical software package for quantitative CBCT imaging and evaluate its performance on three pelvis patients. By providing high-quality CBCT images with a high efficiency, our software shows promises on increasing the accuracy of advanced CBCT-based clinical applications for IGRT, such as dose calculation and adaptive radiation therapy.
Funding Support, Disclosures, and Conflict of Interest: NIH