Improvements in HN CBCT Image Quality Using a Novel Iterative Reconstruction Algorithm
Gardner S, Mao W, Liu C, Aref I, Mohamed E, Lee J, Pradhan D, Siddiqui F, Movsas B, and Chetty I. Improvements in HN CBCT Image Quality Using a Novel Iterative Reconstruction Algorithm. J Med Phys 2019; 46(6):e152-e153.
J Med Phys
Purpose: To use quantitative (contouring variability) and qualitative (image quality) means to evaluate clinical utility of a novel, commercially available iterative cone-beam CT (CBCT) reconstruction for imaging of HN cancer patients. Methods: Ten HN cancer patients were selected for this study. For both aspects of the study (contouring and image quality), observers were blinded to the reconstruction algorithm used. For each patient, iterative CBCT(iterative-CBCT) was compared to conventional CBCT(FDK-CBCT) using contouring and image quality analysis. For contouring analysis, 3 experts contoured parotids, submandibular glands, spinal canal, and brainstem on both image sets. Consensus contours were generated in CERR using STAPLE method for use as reference standard. The following metrics were used to compare observer contours to consensus: Dice coefficient and Hausdorff distance. The iterative-CBCT data was compared to FDK-CBCT data using t-test(P < 0.05-significant). For image quality analysis, 11 observers graded image sets on scale ranging from:1(indicating iterative-CBCT image quality is far superior to FDK-CBCT) to 5(indicating iterative-CBCT is far inferior to FDK-CBCT). Results: The most notable improvement in contouring was found for parotid glands. Parotid contours on iterative-CBCT displayed 2% improvement in Dice coefficient (P = 0.03) and 2 mm improvement in Hausdorff distance (P < 0.01). No statistically significant differences were found for submandibular glands, spinal canal, or brainstem contours. For HN images, the overall average[st.dev] observer score was 2.40[0.88], indicating that the iCBCT image sets displayed improved image quality. Observers noted equivalent or improved image quality for 99/110 (90%) image evaluations and far superior image quality for 14/110(12.7%) image evaluations. Conclusion: Observers noted an improvement in image uniformity, noise level, and overall image quality for CBCT images generated using a novel iterative reconstruction algorithm. Expert observers displayed improvement in ability to consistently delineate parotid glands. Thus, the novel iterative reconstruction algorithm analyzed in this study is capable of improving image quality and soft tissue visualization for HN cancer RT.