Olga Gopan1, Qiuwen Wu. 1. Department of Radiation Oncology, Wayne State University, Detroit, Michigan, USA.
Abstract
PURPOSE: To evaluate the accuracy of three-dimensional (3D) surface imaging system (AlignRT) registration algorithms for head-and-neck cancer patient setup during radiotherapy. METHODS AND MATERIALS: Eleven patients, each undergoing six repeated weekly helical computed tomography (CT) scans during treatment course (total 77 CTs including planning CT), were included in the study. Patient surface images used in AlignRT registration were not captured by the 3D cameras; instead, they were derived from skin contours from these CTs, thereby eliminating issues with immobilization masks. The results from surface registrations in AlignRT based on CT skin contours were compared to those based on bony anatomy registrations in Pinnacle(3), which was considered the gold standard. Both rigid and nonrigid types of setup errors were analyzed, and the effect of tumor shrinkage was investigated. RESULTS: The maximum registration errors in AlignRT were 0.2° for rotations and 0.7 mm for translations in all directions. The rigid alignment accuracy in the head region when applied to actual patient data was 1.1°, 0.8°, and 2.2° in rotation and 4.5, 2.7, and 2.4 mm in translation along the vertical, longitudinal, and lateral axes at 90% confidence level. The accuracy was affected by the patient's weight loss during treatment course, which was patient specific. Selectively choosing surface regions improved registration accuracy. The discrepancy for nonrigid registration was much larger at 1.9°, 2.4°, and 4.5° and 10.1, 11.9, and 6.9 mm at 90% confidence level. CONCLUSIONS: The 3D surface imaging system is capable of detecting rigid setup errors with good accuracy for head-and-neck cancer. Further investigations are needed to improve the accuracy in detecting nonrigid setup errors.
PURPOSE: To evaluate the accuracy of three-dimensional (3D) surface imaging system (AlignRT) registration algorithms for head-and-neck cancerpatient setup during radiotherapy. METHODS AND MATERIALS: Eleven patients, each undergoing six repeated weekly helical computed tomography (CT) scans during treatment course (total 77 CTs including planning CT), were included in the study. Patient surface images used in AlignRT registration were not captured by the 3D cameras; instead, they were derived from skin contours from these CTs, thereby eliminating issues with immobilization masks. The results from surface registrations in AlignRT based on CT skin contours were compared to those based on bony anatomy registrations in Pinnacle(3), which was considered the gold standard. Both rigid and nonrigid types of setup errors were analyzed, and the effect of tumor shrinkage was investigated. RESULTS: The maximum registration errors in AlignRT were 0.2° for rotations and 0.7 mm for translations in all directions. The rigid alignment accuracy in the head region when applied to actual patient data was 1.1°, 0.8°, and 2.2° in rotation and 4.5, 2.7, and 2.4 mm in translation along the vertical, longitudinal, and lateral axes at 90% confidence level. The accuracy was affected by the patient's weight loss during treatment course, which was patient specific. Selectively choosing surface regions improved registration accuracy. The discrepancy for nonrigid registration was much larger at 1.9°, 2.4°, and 4.5° and 10.1, 11.9, and 6.9 mm at 90% confidence level. CONCLUSIONS: The 3D surface imaging system is capable of detecting rigid setup errors with good accuracy for head-and-neck cancer. Further investigations are needed to improve the accuracy in detecting nonrigid setup errors.
Authors: Sang Kyu Lee; Sheng Huang; Lei Zhang; Ase M Ballangrud; Michalis Aristophanous; Laura I Cervino Arriba; Guang Li Journal: J Appl Clin Med Phys Date: 2021-03-31 Impact factor: 2.102