7T-Like T1-Weighted and TOF MRI Synthesis from 3T MRI with Multi-contrast Complementary Deep Learning

Zheng Zhang, Zechen Zhou, Lei Xiang, Kelei He, Zhiqing Zhu, Xingang Wang, Zhiming Zeng, Hongqin Liang, Chen Liu

Abstract: Ultra-high-field (7T) MR imaging offers superior resolution and exceptional anatomical details when compared to conventional 3T MRI. However, the current scarcity and higher cost of 7T MRI scanners limit their accessibility in both clinical and research field. This study introduces a novel approach for simultaneously synthesizing 7T T1-weighted and Time-of-Flight (TOF) Magnetic Resonance Angiography (MRA) images from corresponding 3T T1-weighted and TOF MR scans. This method emphasizes vessel restoration using a unique loss function that integrates multi-view Maximum Intensity Projection (MIP) of the TOF image volume. The proposed method leverages complementary information between 3D T1-weighted and TOF images to enhance 3T images achieving 7T-like image quality. Extensive experiments demonstrate its superior performance compared to existing single-/multi-contrast image reconstruction methods, highlighting the significance of multi-view MIP loss and multi-contrast restoration in improving clinical 3T MRI scans.