Deep Learning-Based TOF-MRA Generation Using a Single mGRE Sequence for Enhanced Peripheral and Large Vessel Visualization: A Feasibility Study

Motivation: This study demonstrates the potential of generating high-quality vascular images from a single mGRE sequence, leveraging both inflow and susceptibility information to visualize veins and arteries without additional TOF-MRA scans. Goal(s): To develop a GAN-based deep learning framework with a U-Net generator that creates TOF-MRA-like images from mGRE data, improving visualization of both small and large vessels without the need for additional scans. Approach: We processed T2*-corrected and uncorrected mGRE data and evaluated model performance through qualitative assessment of maximum intensity projection images and quantitative metrics. Results: Deep learning-generated TOF images provided enhanced vein and artery detail, comparable to TOF-MRA quality. Impact: This study utilizes the inflow and susceptibility information inherent in mGRE to generate enhanced TOF images that visualize both veins and arteries through a deep learning generative model, advancing vascular imaging without additional TOF-MRA scans.