Constrained model-predictive thermal dose control for MRI-guided ultrasound thermal treatments

Model identification and automatic thermal dose feedback control techniques that we have developed for MR-thermometry based ultrasound thermal treatments are described and experimentally evaluated. Pre-treatment, low-power ultrasound sonications, and the resulting temperature elevations were measured using fast MR-thermometry, and used to obtain the specific absorption rate and effective perfusion distributions in the phantoms and an in-vivo canine thigh. The identified thermal and ultrasound models were used in our previously developed constrained model predictive controller to automatically and explicitly realize the efficacy and safety objectives of thermal therapies. In the proposed control approach, the treatment efficacy is characterized by the thermal dose delivered to the target, and the safety objective is formulated in terms of the user-specified maximum allowable temperatures in selected normal tissue locations. Phantom and in-vivo subjects, sonicated with a fixed, focused ultrasound transducer, were used to experimentally evaluate the performance and robustness of the modelling and control techniques. The results demonstrate that the presented model predictive controller delivers the desired thermal dose to the target while limiting the temperatures in the selected normal tissue locations below the specified value. The successful implementation of the MRI-based modelling and control techniques for the simplified stationary transducer cases demonstrate the potential of the proposed approach for advanced and clinically realistic cases that involve scanning ultrasound energy fields. Preliminary simulation results are presented extending the control technique to the moving transducer case.