Validation of MRI to TRUS registration for high-dose-rate prostate brachytherapy

PURPOSE: The objective of this study was to develop and validate an open-source module for MRI to transrectal ultrasound (TRUS) registration to support tumor-targeted prostate brachytherapy. METHODS AND MATERIALS: In this study, 15 patients with prostate cancer lesions visible on multiparametric MRI were selected for the validation. T2-weighted images with 1-mm isotropic voxel size and diffusion weighted images were acquired on a 1.5T Siemens imager. Three-dimensional (3D) TRUS images with 0.5-mm slice thickness were acquired. The investigated registration module was incorporated in the open-source 3D Slicer platform, which can compute rigid and deformable transformations. An extension of 3D Slicer, SlicerRT, allows import of and export to DICOM-RT formats. For validation, similarity indices, prostate volumes, and centroid positions were determined in addition to registration errors for common 3D points identified by an experienced radiation oncologist. RESULTS: The average time to compute the registration was 35 +/- 3 s. For the rigid and deform able registration, respectively, Dice similarity coefficients were 0.87 +/- 0.05 and 0.93 +/- 0.01 while the 95% Hausdorff distances were 4.2 +/- 1.0 and 2.2 +/- 0.3 mm. MRI volumes obtained after the rigid and deformable registration were not statistically different (p > 0.05) from reference TRUS volumes. For the rigid and deformable registration, respectively, 3D distance errors between reference and registered centroid positions were 2.1 +/- 1.0 and 0.4 +/- 0.1 mm while registration errors between common points were 3.5 +/- 3.2 and 2.3 +/- 1.1 mm. Deformable registration was found significantly better (p < 0.05) than rigid registration for all parameters. CONCLUSIONS: An open-source MRI to TRUS registration platform was validated for integration in the brachytherapy workflow. (C) 2017 American Brachytherapy Society. Published by Elsevier Inc. All rights reserved.