Fast and Non-Invasive Oxygen Extraction Fraction Mapping Using Hyperoxic BOLD-fMRI and Deep-Learning-Accelerated Relaxometry

Motivation: Oxygen extraction fraction (OEF) mapping by mqBOLD MRI requires contrast-agent-based DSC MRI to quantify cerebral blood volume (CBV) and suffers from overall long acquisition times. Goal(s): To investigate the utility of CBV estimates from non-invasive, hyperoxic BOLD-fMRI together with deep-learning-(DL-)accelerated R2ʹ measurements for OEF mapping. Approach: Compare CBV from hyperoxic BOLD-fMRI and two different models (Bulte vs. Blockley), R2ʹ from DL-accelerated vs. non-accelerated acquisition, and the respective combinations of both parameters for OEF mapping. Results: Hyperoxic BOLD-fMRI enables non-invasive CBV quantification, and DL-acceleration reduces R2ʹ acquisition times by ~82%. Systematic bias in resulting OEF maps needs further investigation. Impact: We investigated CBV estimation from hyperoxic BOLD-fMRI with two quantification models as a non-invasive alternative to contrast-agent-based DSC MRI, and the reduction of R2ʹ acquisition time by DL-based acceleration. Ultimately, this parameter combination enables fast and non-invasive OEF mapping.