2956 - Serial Characterization and Stability of Quantitative MRI at the Voxel Level during Prostate Radiotherapy

Purpose/Objective(s): Quantitative MRI (qMRI) is emerging as a powerful technique for detecting biological changes in tissue microenvironments, such as those occurring in the planning target volume (PTV) during radiotherapy (RT). Prior longitudinal studies, particularly in head and neck cancer, have demonstrated localized changes in diffusion- and relaxation-based qMRI parameters during RT, as well as relative stability in the spatial distribution of qMRI-defined subregions over time. Whether these phenomena generalize across disease sites is unknown. This study investigates both (i) the longitudinal evolution of qMRI parameters, and (ii) the temporal stability and spatial coherence of qMRI parameters in prostate cancer. This work supports a foundation for cross-site validation of qMRI biomarkers and their potential integration into biologically guided adaptive RT workflows.

Materials/Methods: Twenty-two prostate cancer patients treated with conventionally fractionated MR-guided RT (70 Gy in 28 fractions) were included in this study. Serial intravoxel incoherent motion, multiple flip angle SPGR, and CPMG sequences were acquired on a 1.5T MR-Linac. ADC, T₂, f, and D (derived from b-values 150 and 550 s/mm²) maps were generated and registered to a common coordinate space. Voxelwise parameter estimates were extracted from both the prostate contour and a reference contour drawn in the gluteus maximus at the level of the prostate. Mean and median values for the whole prostate ROI were compared between the first and final fractions of RT using a paired t-test; linear regression was fit using multiple mid-RT time points and compared to zero using a two-sided t-test. Voxelwise response to RT was evaluated using custom scripts to visualize change across RT, slope, and RMSE.

Results: Whole prostate ADC, D, f, and T₂ increased across the course of RT. T₂ and f increased significantly (p-values 0.0166, 0.0074 respectively); ADC and D also increased, but not significantly (p-values of 0.955, 0.183 respectively). No significant changes were observed in the control contour for any qMRI parameter. Voxelwise reconstructions of the prostate revealed spatial patterns of response in T₂ and f, with the greatest degree of change occurring in the periphery of the gland. Reconstructions of ADC and D had less spatial coherence to the changes occurring during RT.

Conclusion: Serial qMRI across the length of RT identified several responses to RT consistent with prior literature, including increases in diffusion, perfusion, and T₂. These changes were most pronounced in the periphery of the prostate, indicating regional patterning of the qMRI response to RT. These findings echo those seen in other disease sites such as the head and neck. Voxelwise monitoring of qMRI maps during RT may support biologically guided adaptive RT across disease sites.