SU-E-J-45: Validation of the ExacTrac Virtual Isocenter Based Target Localization Method

With no stable landmarks available for localization, a 'virtual isocenter' "'or surrogate landmark near the target'" can be used for image guidance. However, using a virtual isocenter in ExacTrac has not been thoroughly validated. This study evaluates its target localization accuracy and investigates the impact of two different couch correction sequences.

A CT scan was acquired on an anthropomorphic thoracic phantom with a 2mm-diameter ball bearing (BB) marker implanted in thelung region. A treatment plan was created with isocenter placed at the BB center, and exported to ExacTrac. In ExacTrac, a virtual isocenter wasplaced on a spine vertebral body where three translational shifts (8.8cm laterally, 1.5cm longitudinally and 6cm vertically) were present. A series ofcouch rotations (+/-3 degrees, 1 degree increment) was intentionally applied to simulate angular setup variations. For each rotation, two stereoscopic x-rayimages were acquired and fused using the ExacTrac 6D registrationalgorithm. Calculated shifts were applied using two sequences: (1)automatic 5D corrections (three translations/two robotic couch rotations) followed by manual couch rotation; (2) manual couch rotation then automatic 5D corrections. After each ExacTrac localization, orthogonal (anterior-posterior and right-lateral) portal images were acquired to quantify BB center deviations from the radiation isocenter as an indicator of residual error.

Minimal difference between investigated table correction sequences was observed. Average translational deviations between the BB and radiation isocenter (mean+/-1SD) were 0.3+/-0.3mm and 1.0+/-0.2mm for lateral and vertical axis respectively. Longitudinally, the deviations were 0.8+/-0.4mm from the anterior-posterior image and 0.1+/-0.3mm from the right-lateral image. The systematic difference (0.7+/-0.1 mm) between thetwo may have been attributed to gantry sagging during rotation.

ExacTrac system successfully corrected angular shifts using the virtual isocenter method in a rigid phantom setup. The sequence ofcouch correction did not influence the localization accuracy. Further patient study is warranted.