Motivation: Magnetic Resonance Fingerprinting (MRF) is a powerful quantitative imaging technique, but noise significantly impacts tissue property estimation accuracy. Goal(s): This study aims to enhance denoising methods for MRF, addressing the critical need for improved image quality in clinical applications. Approach: We developed a denoising process that integrates randomized SVD projections from MRF data into a deep learning model and utilizes inner products to match the denoised data with a standard MRF dictionary for property estimation. Results: Our method achieved substantial noise reduction, with quantitative metrics showing improved inner product measures compared to traditional techniques. Impact: This study provides a novel approach to enhance MRF through advanced denoising techniques, potentially improving diagnostic accuracy and clinical outcomes in quantitative magnetic resonance imaging.
Lo et al. (Tue,) studied this question.