MR-guided high-intensity focused ultrasound (HIFU) treatment of prostate cancer is a promising non-invasive approach. The outcomes of these treatments can be improved with more accurate visualization of ablative necrosis surrounding a tumor. Diffusion-weighted imaging (DWI) could provide valuable information about tissue viability without injection of contrast. However, currently DWI requires high number of excitations (NEX), averages, takes several minutes and is not practical for intraprocedural monitoring. To make DWI a practical tool for monitoring of prostate HIFU, we propose to replace high NEX DWI acquisitions, with NEX=2 acquisitions and subsequent deep-learning denoising of these image.