Magnetic resonance (MR) guided high-intensity focused ultrasound (HIFU) ablation procedures are promising non-invasive alternatives to surgical intervention for localized cancers. A significant challenge for HIFU cancer treatment is non-invasively predicting the final region of non-viable tissue resulting from the HIFU ablation procedures to ensure the entire tumor is treated. Histological analysis of resected tissue is the gold-standard method for identifying non-viable tissue from HIFU ablation procedures. Tissue excised 3-5 days after treatment captures non-viable tissue from immediate and delayed HIFU treatment effects and minimizes tissue clearing from the healing process, ultimately providing an accurate, microscopic representation of the non-viable volume resulting from HIFU ablation. Clinical MR images used to predict the non-viable tissue have been shown to be imprecise versus histological analysis when compared using spatially non-specific measurements, such as volume fractions. However, direct, voxel-wise correlation of MR imaging and histology imaging is challenging because processing tissue for histology destroys the spatial relationship between MR images and histological imaging. Previously developed MR to histology registration methods use methods that are limited in their application to many HIFU ablation applications and introduce significant error in the registration results. The primary objective of this work is to develop MR to histology registration methods for HIFU applications to facilitate volumetric, voxel-wise comparisons between MR images and gold-standard histology non-viable tissue labels. Volume-preserving registration methods were developed to register treatment-day MR images with MR images acquired 3-5 days post-treatment. The registration methods were able to correct for the change in patient pose and orientation with a registration error of 1.33+/-0.16 mm while limiting the total volume change to 0.28+/-0.11 percent. Subsequently, a novel 3D, surface-based registration workflow applicable to HIFU procedures was developed to register post-treatment MR images and histology images with an average error of 1.00+/-0.13 mm. Finally, the composed registration results were used to investigate 3D non-contrast, multi-parametric MR predictions of non-viability that are more accurate than clinical predictions. The registration methods developed in this work will improve acute MR assessment of HIFU ablation procedures and ultimately improve HIFU treatment efficacy and their application to localized cancers.