Dictionary generation for multi-parametric cardiac Magnetic Resonance Fingerprinting (MRF) is a significant bottleneck as subject-specific dictionaries must be created accounting for the subject’s heart rate variability and dictionaries grow exponentially with the number of parameters considered. Here we propose a feedforward neural network to generate cardiac MRF dictionaries for T1, T2 and T1ρ. The proposed approach was tested on simulations and in-vivo data, generating dictionaries in 3 seconds. The proposed method achieved a good match to dictionaries generated with Extended Phase Graph (EPG) simulations with mean relative errors for myocardium T1, T2 and T1ρ ranging from 1.7% to 5.1%.
This abstract and the presentation materials are available to members only; a login is required.