A new regularizer is proposed for the magnitude least‐squares optimization algorithm, to ensure robust parallel transmit RF shimming and small‐tip‐angle multispoke pulse designs for ultrahigh‐field MRI.

A finite‐difference regularization term is activated as an additional regularizer in the iterative magnitude‐least‐squares based pulse design algorithm when an unwanted flip angle null distribution is detected. Both simulated and experimental maps from different transmit arrays and different human subjects at 7 T were used to evaluate the proposed algorithm. The algorithm was further demonstrated in experiment with dynamic multislice RF shimming for a single‐shot gradient‐echo EPI for human functional MRI at 7 T.

The proposed finite‐difference regularizer effectively prevented excitation null to be formed for RF shimming and small‐tip‐angle multispoke pulses, and improved the latter with a monotonic trade‐off relationship between flip angle error and RF power. The proposed algorithm was demonstrated to be effective with several head‐array geometries by simulation and with a commercial head array with 12 healthy human subjects by experiment. During a functional MRI scan at 7 T with dynamic RF shimming, the proposed algorithm ensured high image SNR throughout the human brain, compared with near‐complete local signal loss by the conventional magnitude‐least‐squares algorithm.

Using finite‐difference regularization to avoid unwanted solutions, the robustness of RF shimming and small‐tip‐angle multispoke pulse design algorithms are improved, with better flip angle homogeneity and a monotonic trade‐off relationship between flip angle error and RF power.