Technique for Accurate GTV Definition in MR-Guided HDR Prostate Brachytherapy

Purpose: Despite a strong rationale and enthusiasm for distinguishing regions of tumor-dense burden within the prostate gland for specific dose delivery, methods and tools for clinical investigation have not yet matured. A technique is presented for integrating pathologically referenced multiparametric MRI using deformable registration for HDR brachytherapy planning.Materials and Methods: Four patients enrolled on a prospective trial of tumor-targeted MRI-guided HDR salvage brachytherapy were evaluated, each receiving two fractions over 10 days. Multiparametric online MRI-guided prostate biopsy with 3D image verification of needle core locations enabled cores and suspicious tumor volumes to be mapped onto T2 MRI. The GTV was defined as the shared boundary of suspicious image features and malignant biopsy cores. On treatment day, radiation oncologist estimated the intended GTV boundary based on common anatomical MRI landmarks. Inverse treatment planning utilized IPSA objectives to achieve tumor V11Gy>95% and prostate V8Gy>95%, while respecting OAR constraints. MORFEUS, a biomechanical model-based deformable image registration technique, was used to perform deformable registration of diagnostic images acquired during biopsy onto the brachytherapy planning image. Common 3D points in the prostate were identified to determine the absolute error between the observed and MORFEUS-predicted point displacements. The registered GTV was compared with the estimated GTV using Dice's coefficient and tumor D98 and V11Gy.Conclusions: Feasibility of integrating pathologically referenced multiparametric MRI for GTV delineation during HDR prostate brachytherapy has been demonstrated. The accuracy achieved with the deformable registration was smaller than the largest voxel dimension. A small, but potentially relevant dosimetric discrepancy between estimated and deformably registered GTV was observed. Ongoing work will explore strategies to improve AP registration accuracy, and streamline procedures to enable online use. Purpose: Despite a strong rationale and enthusiasm for distinguishing regions of tumor-dense burden within the prostate gland for specific dose delivery, methods and tools for clinical investigation have not yet matured. A technique is presented for integrating pathologically referenced multiparametric MRI using deformable registration for HDR brachytherapy planning. Materials and Methods: Four patients enrolled on a prospective trial of tumor-targeted MRI-guided HDR salvage brachytherapy were evaluated, each receiving two fractions over 10 days. Multiparametric online MRI-guided prostate biopsy with 3D image verification of needle core locations enabled cores and suspicious tumor volumes to be mapped onto T2 MRI. The GTV was defined as the shared boundary of suspicious image features and malignant biopsy cores. On treatment day, radiation oncologist estimated the intended GTV boundary based on common anatomical MRI landmarks. Inverse treatment planning utilized IPSA objectives to achieve tumor V11Gy>95% and prostate V8Gy>95%, while respecting OAR constraints. MORFEUS, a biomechanical model-based deformable image registration technique, was used to perform deformable registration of diagnostic images acquired during biopsy onto the brachytherapy planning image. Common 3D points in the prostate were identified to determine the absolute error between the observed and MORFEUS-predicted point displacements. The registered GTV was compared with the estimated GTV using Dice's coefficient and tumor D98 and V11Gy. Conclusions: Feasibility of integrating pathologically referenced multiparametric MRI for GTV delineation during HDR prostate brachytherapy has been demonstrated. The accuracy achieved with the deformable registration was smaller than the largest voxel dimension. A small, but potentially relevant dosimetric discrepancy between estimated and deformably registered GTV was observed. Ongoing work will explore strategies to improve AP registration accuracy, and streamline procedures to enable online use.