Accuracy of a 3D laser∕camera surface imaging system for setup verification of the pelvic and thoracic regions in radiotherapy treatments.

PURPOSE The aim of this paper is to investigate the accuracy of a laser∕camera surface imaging system (Sentinel) when used on patients treated in the thorax or pelvic regions and to evaluate system employment conditions and patient setup procedures that provide more accurate results. METHODS The system was tested on two groups of patients for whom different patient setup procedures and Sentinel employment conditions were considered. For the first group of 33 patients (FG) no changes to the usual setup procedures were made and a surface extension limited to the treated region was considered. For the second group of 14 patients (SG) the reproducibility of external body surfaces, including body parts not in close proximity to the treatment site, was optimized and a wider surface was captured. In all cases the system accuracy was evaluated comparing registration results from concurrent Sentinel and cone beam CT (CBCT) acquisitions for a total of 192 occasions. External body surfaces, extracted from planning CT studies, were used as reference in both cases, but for SG also surface data captured by Sentinel system at the first treatment were employed. RESULTS In the 90th percentile of the distributions reporting CBCT and Sentinel registration parameters, absolute differences for FG were less than 6.4 mm and 3.8°. Better performances were observed for SG (≤5.7 mm and 2°). Mean absolute differences between three translation and three rotation parameters of CBCT and Sentinel were: less than 3.5 mm and 2.1° and 3.7 mm and 1.3° in FG for thorax and pelvis, respectively, and less than 2.8 mm and 1° and 2.7 mm and 0.9° for pelvis and thorax, respectively, in SG. No advantage in considering surface data captured by Sentinel as a reference instead of the surface extracted from the planning CT was observed. CONCLUSIONS The accuracy of Sentinel system in detecting errors is influenced by the extension and reliability of the surface used. When the reproducibility of external body surfaces was optimized differences between CBCT and Sentinel registration parameters resulted less than 5.7 mm and 2° in the 90% of the pelvis and thorax considered cases. No advantage in considering a Sentinel acquisition as reference was observed.