Denoising Diffusion Probabilistic Models for Magnetic Resonance Fingerprinting

To reduce scan time, approach to quantitative MRI, enabling the mapping of multiple MRF uses short-length acquisition sequences to encode multiple tissue properties from a single, accelerated scan. However, tissue properties simultaneously and applies compressed achieving accurate reconstructions remains challenging, sensing to subsample only a fraction of the spatiotemporal particularly in highly accelerated and undersampled acquisitions, k-space data. However, faster scans lead to challenges in image which are crucial for reducing scan times. While deep reconstruction, including aliasing artifacts from k-space learning techniques have advanced image reconstruction, the undersampling and limited tissue property information due recent introduction of diffusion models offers new possibilities to the truncated acquisition sequences. Effective image reconstruction for imaging tasks, though their application in the medical algorithms are needed to tackle these challenges field is still emerging. Notably, diffusion models have not yet and improve the accuracy and precision of tissue parameter been explored for the MRF problem. In this work, we propose estimation.