SU‐E‐J‐205: Fast Computation of Actual Dose Delivered to the Patient for IMRT Treatment Quality Assessment and Adaptive Radiotherapy

Abstract

Purpose: To fast and accurately compute the actual dose delivered to the patient during the intensity‐modulated radiotherapy (IMRT) for treatment quality assessment and adaptive radiotherapy. Methods: A fast deformable image registration (DIR) and dose computation engines both accelerated by a graphics processing unit (GPU) are implemented and integrated into a unified framework. DIR automatically aligns the planning CT (pCT) to daily cone‐beam CTs (CBCTs) by a rigid registration followed by a b‐spline registration using mutual information as a similarity metric. The dose engine computes the dose on the registered pCT by using a superposition/convolution technique. The implemented tool was tested on a retrospective image data set of a head and neck cancer patient. The pCT was registered to 13 CBCTs (among 30 fractions) and the actual dose was computed on each registered CT. Results: The average computation time for DIR of the pCT and CBCT of 512×512×136 was less than one minute. The dose was computed on a 171×171×136 grid with the grid size of 3.8×3.8×3.0 mm3, and the average computation time was 0.6 second/beam. The registered pCT showed 7.5% improvement in terms of normalized cross‐correlation with the CBCT while simple couch shifts showed only 4.4% improvement. Our GPU dose engine enables over 30 times speed‐up over the highly optimized commercial treatment planning software (Pinnacle, Philips, Madison, WI). Conclusion: A unified framework for computing the actual dose delivered to the patient was implemented. The implemented tool significantly improves the accuracy of dose computation by DIR followed by direct dose computation on the deformed pCT while maintaining a substantially improved performance over CPU‐based implementations. The superior performance of this tool will enable an efficient and accurate on‐line dose computation for treatment quality assessment and adaptive radiotherapy. This work was supported in part by National Science Foundation under grant No. EEC9731748, in part by Johns Hopkins University internal funds, and in part by National Institute of Health ‐ National Cancer Institute under grant No. 2R42CA137886.