PURPOSE: Online optimization of annular-phased-array hyperthermia (HT) is based on planning tools and magnetic resonance (MR) thermometry. Until now, the method has been validated in phantoms. Further developments and extensions are required for clinical purposes. In particular, the problem of deducing the electric field distribution inside the patient from MR thermometry is ill-posed, which leads to an amplification of measurement errors. A method to overcome this difficulty is proposed. METHODS: The authors utilized a regularized Gauss-Newton algorithm with a fast bioheat transfer equation (BHTE) approximation to identify the field parameters. To evaluate the method, simulations with patient models are conducted and a treatment data set obtained from a heat treatment performed in the hybrid HT-MR system at the Charité Medical School is used to visualize the error amplification. RESULTS: The regularization leads to a significantly improved accuracy of the predicted electric fields and temperatures compared to an unregularized approach. The BHTE approximation enables highly accurate temperature predictions in real-time. CONCLUSIONS: Regularization proves to be necessary to identify electromagnetic field parameters. The proposed method is able to reproduce measurements without overfitting to the noise in the MR measurements and results in an improved treatment planning.
PURPOSE: Online optimization of annular-phased-array hyperthermia (HT) is based on planning tools and magnetic resonance (MR) thermometry. Until now, the method has been validated in phantoms. Further developments and extensions are required for clinical purposes. In particular, the problem of deducing the electric field distribution inside the patient from MR thermometry is ill-posed, which leads to an amplification of measurement errors. A method to overcome this difficulty is proposed. METHODS: The authors utilized a regularized Gauss-Newton algorithm with a fast bioheat transfer equation (BHTE) approximation to identify the field parameters. To evaluate the method, simulations with patient models are conducted and a treatment data set obtained from a heat treatment performed in the hybrid HT-MR system at the Charité Medical School is used to visualize the error amplification. RESULTS: The regularization leads to a significantly improved accuracy of the predicted electric fields and temperatures compared to an unregularized approach. The BHTE approximation enables highly accurate temperature predictions in real-time. CONCLUSIONS: Regularization proves to be necessary to identify electromagnetic field parameters. The proposed method is able to reproduce measurements without overfitting to the noise in the MR measurements and results in an improved treatment planning.