C. Velten1,2, M. Gjini2, R. Kabarriti2, N. P. Brodin1, and W. A. Tome2,3; 1Institute for Onco-Physics, Albert Einstein College of Medicine, Bronx, NY, 2Department of Radiation Oncology, Montefiore Medical Center, Bronx, NY, 3Institute for Onco-Physics, Albert Einstein College of Medicine, Bronx, NY
Purpose/Objective(s): To develop functionally guided normal-tissue complication probability (NTCP) modeling for patients undergoing stereotactic body radiotherapy (SBRT) for hepatocellular carcinoma (HCC). Materials/
Methods: Voxelized contrast kinetic modelling was performed on pre-treatment DCE-MRI employing gadoxetate disodium contrast to create volumetric maps correlated with liver function. Modelling of hepatic NTCP was performed using a local damage parallel architecture model with ALBI grade increase as endpoint. Each voxel was assigned a damage probability (logistic function; parameters: EQD50, k). A weighted organ-average fraction of damaged subunits is calculated using the functional metric’s cumulative distribution function as weights. NTCPRTfrom SBRT is calculated using a shifted error function (assuming normal distribution of functional reserve; parameters: mu, sigma) and combined with baseline NTCP0. Volumetric maps of function were employed to perform functionally avoidant SBRT photon and proton planning to evaluate potential NTCP reductions. Results: NTCP modelling was performed using retrospective data from 68 patients. Local damage model parameters were obtained to be EQD50=17Gy2and k=1.68, while NTCP parameters were µ=0.47, s=0.12, and NTCP0=13%. The total number of observed ALBI grade increases were 11/68 patients (16%). 63 patients (93%) had fdmg?[0,0.3) with nine NTCP events (14.2%). The remaining 5 patients had fdmg?[0.3,0.5) with two NTCP events (40%). For comparison, NTCP modelling was performed without functional weighting of the local damage probabilities when calculating fdmg yielding EQD50=19Gy2and k=1.07, µ=0.53, s=0.14, and NTCP0=15%. The highest observed NTCP was 25% and no patient had fdmg=0.4 without functional weighting. Photon and proton SBRT reference and avoidant plans were generated for six patients who had undergone CT simulation followed by same-day DCE-MRI in treatment position. For three patients the predicted NTCP was approximately 13.5% in both reference and avoidant photon and proton plans. For two patients, predicted photon NTCP was higher with 17.8% and 18.3% in reference plans which was reduced to 16.6% and 18.0% in avoidant plans, whereas proton NTCP was predicted to be about 13.8% to 14.1%. Finally, one patient had a predicted photon NTCP of 40.2% in the reference plan compared to 34.5% in the avoidant plan, while the proton reference plan had a predicted 18.3% NTCP which was reduced to 16.5% with functional avoidance. Conclusion: NTCP modelling using functional weighting of functional subunit damage probabilities has been shown to be superior in predicting ALBI grade increases after SBRT for HCC compared to function-agnostic modeling. 3D maps of surrogate quantities for liver function can be used to support decision-making for proton vs. photon SBRT and can guide proton aperture placement and photon SBRT optimization.
