Transmit uniformity and SAR optimization by a deep-learning method in UHF imaging

Motivation: UHF imaging is limited by both transmit uniformity and local SAR. Information on RF electric field is unavailable in conventional MR scan procedure. Goal(s): This work aims to provide real-time RF electric field for joint optimization of imaging uniformity and peak local SAR. Approach: A deep-learning method is proposed to predict the real-time EM field distribution using B1+ data obtained in routine prescan of the imaging procedure. The output field data is used in combined optimization of transmit uniformity and local SAR. Results: In the torso imaging case, this method achieved both improvement of transmit field uniformity and reduction of peak local SAR. Impact: This work studied the feasibility of machine-learning methods for RF field estimation and simultaneous optimization of transmit homogeneity and peak local SAR, aiming to reduce the estimation error of local SAR and increase the available maximum B1.