SynthOEF: Oxygen Extraction Fraction (OEF) Mapping from multi Gradient-Echo Data using a U-Net trained with Synthetically Generated Data

Motivation: Gold standard for OEF measurement is 15O-radiotracer PET which is invasive and not widely available. Previously proposed MRI-based OEF estimation techniques lack microstructural modeling and neglect flow effects. Goal(s): This study aims to develop a detailed signal modeling-based MRI approach combined with deep learning for accurate OEF mapping. Approach: SynthOEF is proposed to address the mentioned weaknesses of the current techniques employing a numerical microstructural signal simulation approach for labeled synthetic data generation. Results: Synthetic and in vivo results show the feasibility of accurate OEF mapping via the proposed approach. OEF changes in stroke lesion and pathologic tissues in AD were successfully visualized. Impact: Oxygen Extraction Fraction (OEF) is an important clinical biomarker of tissue viability. Noninvasive measurement of OEF might have high clinical impact by providing critical knowledge of tissue health in diseases such as stroke, Alzheimer's disease, and multiple sclerosis (MS).