F. Bayat1, U. Pyakurel1, R. Sabounchi2, R. C. Bliley1, C. M. Fisher3, R. A. Rabinovitch3, D. C. Binder4, B. D. Kavanagh5, R. M. Lanning5, T. P. Robin5, and C. Altunbas6; 1University of Colorado School of Medicine, Aurora, CO, 2University of Colorado Denver, Denver, CO, 3University of Colorado, Aurora, CO, 4University of Colorado - Denver, Aurora, CO, United States, 5Department of Radiation Oncology, University of Colorado School of Medicine, Aurora, CO, 6University of Colorado, AURORA, CO
Purpose/Objective(s): Cone beam computerized tomography (CBCT) image quality plays an important role in target localization, response assessment, and treatment plan adaptation. However, standard CBCT images have lower quality than planning CT (pCT) scans, in part due to scattered X-rays, limiting their utility. To address this, we developed a novel quantitative CBCT (qCBCT) pipeline, featuring a 2-dimensional anti-scatter grid (2D ASG) and associated data processing methods. This interim study hypothesizes that qCBCT demonstrates improved image quality compared to standard CBCT, as measured by established image quality metrics. Materials/
Methods: To date, 17 patients with cancers of the abdomen and pelvis were enrolled into this IRB-approved trial. Each participant underwent CBCT imaging twice within 30 minutes (pelvis CBCT protocol in the CBCT-linac as well as qCBCT protocol employing a prototype 2D ASG developed by our group). CBCT scan parameters, patient immobilization, and imaging dose were the same in both scans. Standard clinical CBCT images were reconstructed using both standard and advanced Iterative Reconstruction (IR) options in the CBCT-linac. The qCBCT raw projections were first processed offline using a data correction pipeline with in-house developed software and reconstructed by employing filtered back projection and Projection onto Convex Sets (POCS) based IR algorithms. Image quality metrics - structural similarity index (SSIM) and CT number accuracy with respect to the gold-standard pCT images, CT number nonuniformity, and Contrast to Noise Ratio (CNR) - were calculated for all CBCT images. In addition, trends in image quality metrics as a function of patients’ average separation in AP and lateral directions were evaluated. Results: Clinical CBCT and qCBCT images exhibited an average SSIM of 0.77±0.1 and 0.79±0.09, respectively, whereas average CT number accuracy of clinical CBCT and qCBCT was 39±43 HU and 18±17 HU, respectively. Additionally, CT nonuniformity in clinical CBCT scans and qCBCT was 35±41HU and 18±17 HU, respectively. Soft tissue CNR in both CBCT methods were comparable, 7±4 and 7±3. While image quality metrics in clinical CBCT and qCBCT images were more comparable in patients with lateral separations less than 35 cm, clinical CBCT image quality metrics degraded faster for separations larger than 35 cm. Conclusion: While structural similarity and CNR were comparable in clinical CBCT and qCBCT images, both CT number accuracy and nonuniformity were improved substantially with the qCBCT method. The use of a 2D anti-scatter grid with qCBCT did not have any adverse impact on soft tissue contrast and appearance. Lastly, patients with larger AP and lateral separations may benefit more from qCBCT’s improved quality when compared to the current clinical CBCT.
