Objective: The motivation of this study is to establish a procedure for facilitating real-time Magnetic Resonance (MR) guided temperature steering and focusing.

Methods: By feeding back simulated MR temperature images at time points during a finite-difference simulated heating procedure, an iterative least-squares algorithm was used to progressively construct a model governing the relationship between antenna parameters (amplitudes, phases) and the temperature distribution. At each point in the iterative procedure, antenna parameters that were optimized using the model at hand (optimized to maximize tumor heating) were applied as heating inputs for the next iteration. The optimal antenna driving vector was determined as the maximal eigenvector of the averaged tumor temperature matrix in the model. Using this iterative correction procedure, the driving vector that focuses power to maximize tumor temperature is determined in n² steps (in the absence of noise in the MR images), for an n-antenna system. The algorithm was validated using a human thigh tumor model (Figure 1). A four-antenna cylindrical mini annual phased array was used to heat the thigh (tumor length 10.8 cm). To gauge the algorithm’s robustness, two cases were tested: 1) patient properties such as electric conductivity, electric permittivity, thermal conductivity, density and blood perfusion were erroneously assumed to be ±30% deviated from the actual properties, and 2) the patient position was erroneously assumed to be shifted by 2.0 cm from the actual position.

Results: In both cases, the algorithm successfully converged to focus heating at the tumor (see Figure 2) by (1) correcting for the material properties deviation, and (2) correcting for the position deviation.

Conclusions: The numerical results suggest that the algorithm could provide a robust theoretical basis for MR-guided temperature feedback control.