Application of an adaptive encoding matrix generation algorithm in multi-channel B₁⁺ mapping

To improveB₁⁺mapping accuracy in ultra-high field multi-channel parallel transmission (pTx) MRI, this study introduces an iterative optimization framework for anatomy-specific encoding matrix design. Using an 8-channel pTx body coil on a 5T MRI system, we acquired multi-channelB₁⁺maps and constructed a denoised ground truth via filtering. Gaussian noise was added to simulate realistic acquisition conditions. Single-channelB₁⁺maps were decoded iteratively under newly generated encoding matrices, and optimization was driven by minimizing the normalized root mean square error (NRMSE) between decoded and ground truth maps. After 200 iterations, the optimized matrix achieved an average NRMSE reduction of 38.8% compared to the conventional fixed matrix, with particularly notable suppression of noise-related artifacts in low-SNR regions. All eight channels showed statistically significant improvements (p < 0.05), with NRMSE reduction ranging from 22.4% to 61.4% across channels. The method demonstrates strong potential for anatomy-specific calibration in pTx systems. Future work will prioritize the development of real-time optimization strategies to transition from the current offline processing framework to a clinically feasible online implementation. Additionally, we will validate the method across diverse anatomical regions and field strengths to promote broader clinical adoption.