1106 - Toward a Common Biological Dose Scale for SBRT and SIRT Using Imaging-Derived Liver Functional Toxicity Response
Presenter(s)
Y. Wang1, Z. Lu2, J. Mikell3, D. Polan4, M. P. Aryal4, K. C. Cuneo4, R. K. Ten Haken4, T. S. Lawrence4, Y. Cao4, Y. Dewaraja2, and L. Wei4; 1Department of Nuclear Engineering and Radiological Sciences, University of Michigan, Ann Arbor, MI, 2Department of Radiology, Division of Nuclear Medicine, University of Michigan, Ann Arbor, MI, 3WashU Medicine, Department of Radiation Oncology, St. Louis, MO, 4Department of Radiation Oncology, University of Michigan, Ann Arbor, MI
Purpose/Objective(s): We hypothesized that MRI-derived regional functional loss after stereotactic body radiation therapy (SBRT) and Y-90 selective internal radiation therapy (SIRT) can be described by one sigmoidal dose-response relationship after transforming modality-specific physical dose to a shared biological dose scale. Direct comparison on the physical dose axis is not biologically meaningful due to differences between SBRT and SIRT (e.g., dose rate, temporal delivery). We developed a joint optimization framework to estimate (1) a modality-specific biological dose transformation incorporating subvoxel heterogeneity (EUD-based modeling) and (2) a shared logistic response model linking biological dose to MRI-derived functional loss.
Materials/Methods: Twenty-four patients treated with SBRT and nine treated with Y-90 SIRT on IRB-approved protocols underwent liver dynamic gadoxetic acid-enhanced MRI before and 1 month after treatment. 90Y PET/CT was performed soon after SIRT for dosimetry. Voxel-wise gadoxetic acid uptake rate (k1) maps were quantified as regional hepatic function, co-registered to dose maps and used as response to match SBRT and SIRT. To reduce voxel-level noise, liver voxels were aggregated into dose bins, and mean dose and functional change (liver toxicity) were calculated per bin. SBRT dose was expressed as EQD2. For SIRT, mean absorbed dose was converted to EQD2 using a protraction-aware linear-quadratic model with subvoxel heterogeneity (assumed normal distribution) represented by an EUD term, parameterized by ?=(a, a/ß, µ, c), where a and ß are linear-quadratic coefficients, µ is the sublethal repair rate, and c quantifies subvoxel dose heterogeneity for SIRT. A constrained least-squares optimization jointly estimated ? and a shared 4-parameter logistic (4PL) dose-response curve with parameters (A, B, ?, D50,EQD2), where A and B are lower and upper bounds. Goodness-of-fit of the shared dose-response model was assessed using R2.
Results: The estimated biological dose-conversion parameters were a = 0.076 Gy-1, a/ß = 3.3 Gy, µ = 0.20 h-1, and c = 0.51. The corresponding shared 4-PL in the biological-dose (EQD2) domain yielded D50,EQD2 = 37.1 Gy, ? = 0.64, A = -0.06 mL/100g/min and B = 5.5 mL/100g/min, with R2=0.72 for the pooled shared dose-response fit. Back-transformation of D50,EQD2 to modality-specific physical dose yielded D50 = 28.3 Gy (5 fractions) for SBRT and D50 = 103.5 Gy for SIRT.
Conclusion: A joint biological dose-calibration and shared logistic response framework can harmonize SBRT and SIRT functional dose-response modeling on a common biological effect scale. This approach enables more meaningful cross-modality comparison of regional liver tolerance and may support biologically informed treatment planning, response assessment, and integration of external beam radiotherapy with SIRT or radiopharmaceutical therapies.