Premium
TU‐A‐BRA‐10: Real‐Time Markerless Tumor Tracking with MV Imaging and a Dynamic Multi‐Leaf Collimator (DMLC)
Author(s) -
Rottmann J,
Keall P,
Yue Y,
Berbeco R
Publication year - 2012
Publication title -
medical physics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.473
H-Index - 180
eISSN - 2473-4209
pISSN - 0094-2405
DOI - 10.1118/1.4735878
Subject(s) - imaging phantom , fiducial marker , computer vision , computer science , collimator , artificial intelligence , image guided radiation therapy , frame rate , aperture (computer memory) , multileaf collimator , medical imaging , position (finance) , nuclear medicine , linear particle accelerator , physics , optics , medicine , acoustics , beam (structure) , finance , economics
Purpose: The implementation of a real‐time adaptive therapy system. Automatic soft tissue localization (STiL) is used to drive DMLC adaptation maintaining the position of the dynamic treatment aperture relative to the tumor location during the entire breathing cycle. The STiL component utilizes electronic portal images and operates without the need for fiducial markers. The proposed system has the potential to improve treatment accuracy, dose conformity and sparing of healthy tissue. Methods: The system is implemented and tested on a clinical linear accelerator featuring an electronic portal imaging device (EPID) and a DMLC control system. EPID images are continuously acquired at a frame rate of 12.86 Hz. The STiL component processes the images in real‐time, sending its output ‐ the current tumor position ‐ to the DMLC component, which moves the aperture to that position. Image transfer, tumor position calculation and DMLC motion introduce a time lag between tumor position at acquisition time and at the time the treatment aperture reaches this position. We analyze this latency with a dynamic chest phantom driving a 1D sinusoidal motion (20mm superior‐inferior motion range and 4.5s period). We estimate the resulting average geometric systematic error in a clinical setting by driving the phantom with several patient traces (recorded from fiducial tracking during lung SBRT). Results: The individual geometric errors of the StiL and the DMLC component are each smaller than 1 mm. The overall system latency was found to be 210 ms. The average rms‐error of 11 patients traces (172 beams) was found to be (1.8 ± 0.8) mm with this latency. Conclusions: We have implemented a real‐time adaptive therapy system integrating automatic soft tissue tumor localization with DMLC adaptation of the treatment aperture. The functionality of the combined system was tested successfully and the systemic latency and resulting rms error measured. Varian Medical Systems Inc., NCI grant CA93626