Simulating Brain Gradient-Echo Magnetic Resonance Images through Microstructural Modeling

Quantitative magnetic resonance imaging (qMRI) methods usually suffer from the lack of appropriate or easy-to-implement methods of validation.Here, we developed a detailed framework to synthetically generate realistic brain multi gradient-echo (mGRE) data incorporating biophysical modeling of the MRI signals and microstructure brain tissues.In addition to validation, simulated data can also be utilized for the supervised training of deep learning models for the inverse mapping of the microstructural and physiological distributions.The feasibility of one such parameter that is extremely valuable if can be measured noninvasively, oxygen extraction fraction (OEF), is shown.The simulated and in vivo tests demonstrated high accuracy in the mapping.The developed method and the results reflect the significance of simulation-based computational approaches for answering the clinical needs.