Matthias Gebhardt1, Dirk Diehl2, Elfar Adalsteinsson3, Lawrence L. Wald4, Gabriele Eichfelder5
1Siemens Healthcare, Erlangen, Germany; 2Siemens Corporate Technology, Erlangen, Germany; 3Electrical Engineering and Computer Science, Massachusetts Institute of Technology, Cambridge, MA, United States; 4Martinos Center for Biomedical Imaging, Harvard University, Charlestown, MA, United States; 5Applied Mathematics II, University of Erlangen-Nuremberg, Erlangen, Germany
This work addresses the complexity problem inherent in local SAR estimation for parallel transmission (pTx) and shows that a relatively small subset of carefully selected virtual observation points is adequate for prediction and control of maximum local SAR in pTx. We tested the proposed algorithm to detect local SAR maxima by comparison with an exhaustive search over local SAR distribution in numerical simulations of adult male and female subjects for an 8 channel whole body transmit array. The proposed method of model compression for local SAR successfully captured regions of local SAR maxima, but with dramatically reduced computation cost.