University of California, San Francisco San Francisco, CA
J. Everts1, P. Ramesh2, Y. Yang1, M. Feng1, J. Scholey1, M. Ohliger1, Q. Lyu1, X. Miao3, O. W. Chau1, J. Chen4, Z. Fan4, K. Sheng1, and W. Yang1; 1University of California, San Francisco, San Francisco, CA, 2University of California, Los Angeles, Los Angeles, CA, 3Siemens Medical Solutions USA Inc, Malvern, PA, 4University of Southern California, Los Angeles, CA
Purpose/Objective(s): Recent studies have identified the T1 reduction rate for patients undergoing dynamic gadoxetic acid-enhanced magnetic resonance imaging (MRI) as an imaging biomarker for liver function. In this study, we developed a 4pi treatment planning strategy using T1 reduction maps to guide simultaneous non-coplanar beam orientation optimization (BOO) and fluence map optimization (FMO) for maximal liver function sparing in stereotactic body radiation therapy (SBRT) planning. Materials/
Methods: A previously validated Magnetization-Prepared Golden-angle RAdial Sparse Parallel (MP-GRASP) MR sequence was implemented on the 3T simulator for rapid free-breathing T1 mapping in a 3-min scan. With IRB approval, 28 liver cancer patients underwent both CT and MR simulation on the same day. MP-GRASP was performed pre- and 20-min post-gadoxetic acid injection. Quantitative T1 maps were derived. Voxel-wise T1 reduction rate (k1) map was calculated using k1=(T1pre-T1post)/T1pre. The mean k1 values were grouped by the patients’ liver cirrhosis status, which was used to validate k1 as an imaging biomarker for liver function classification. 10 patients were then retrospectively planned for 4pi-SBRT. Tumor coverage was maintained at PTV D95=50Gy. The liver-PTV structure was defined as the normal liver, with previously treated liver regions included. The less functional liver region included liver voxels with k1 < median(k1) for each patient. Group sparsity-regularized beam orientation optimization, with penalization on the product of k1 and optimized dose, was performed to encourage the selection of 15-20 IMRT beam angles optimal for highly functional liver voxel sparing. Following BOO, FMO was performed again using the k1 regularization term. The penalty threshold was set at 5 Gy in each step to assess functional liver sparing. The Fast Iterative Shrinkage Thresholding Algorithm (FISTA) was used to solve the optimization problem. The mean doses to the liver-PTV were compared with- (4pi-wk1) and without- (4pi-wok1) k1 regularization. Paired, two-tailed T-tests were performed to compare liver sparing. Results: Different patterns of k1 maps and wide ranges of k1 values were observed. k1 differentiated patients with normal liver function from those with cirrhosis (55.7 +/- 7.9 % vs. 42.4 +/- 10.5 %, p=0.0034). 4pi-wk1 planning strategy achieved an average functional liver dose of 7.5Gy +/- 5.2Gy compared to 14.8 +/- 9.2Gy from 4pi-wok1 while keeping the same PTV coverage, a statistically significant functional liver dose reduction of 49.2% with p=.010. Conclusion: Quantitative k1 values were validated as a liver function biomarker. Liver function sparing was substantially improved with 4pi-wk1 plans, showing promising benefits for directly incorporating quantitative T1 reduction rate into liver function avoidance 4pi-SBRT planning.
