Proposal for local SAR safety margin in pediatric neuro-imaging using 7 T MRI and parallel transmission.

Objective.Ultra-high field MRI with parallel transmission (pTx) provides a powerful neuroimaging tool with potential application in pediatrics. The use of pTx, however, necessitates a dedicated local specific absorption rate (SAR) management strategy, able to predict and monitor the peak local SAR (pSAR_10g). In this work, we address the pSAR_10gassessment for an in-house built 7 T 16Tx32Rx pediatric head coil, using the concept of virtual observation points (VOPs) for SAR estimation.Approach. We base our study on full-wave electromagnetic simulations performed on a database of 64 numerical anatomical head models of children aged between 4 and 16 years. We built VOPs on different subsets of this database ofN= 2 up to 30 models, and cross-validated the pSAR_10gprediction using non-intersecting subsets, each containing 30 models. We thereby propose a minimum anatomical safety factor (ASF) to apply to the VOP set to enforce safety, despite intersubject variability. Our analysis relies on the computation of the worst case SAR to VOP-SAR ratio, independent of the pTx RF excitation.Main results.The interpolation model provides that the minimum ASF decreases as1+5.37⋅N-0.75withN. Using all 64 models to build VOPs leads to an estimated ASF of 1.24 when considering the VOP validity for an infinite number of subjects.Significance.We propose a general simulation workflow to guide ASF estimation and quantify the trade-off between the number of numerical models available for VOP construction and the safety factor. The approach would apply to any simulation dataset and any pTx setup.