The problem of controlling noninvasive thermal therapies is formulated as the problem of directly controlling thermal dose of the target. To limit the damage to the surrounding normal tissue, the constraints on the peak allowable temperatures in the selected spacial locations are imposed. The developed controller has a cascade structure with a linear, constrained, model predictive temperature controller in the secondary loop. The temperature controller manipulates the intensity of the ultrasound transducer with saturation constraints, which noninvasively heats the spatially distributed target. The main nonlinear thermal dose controller dynamically generates the reference temperature trajectories for the temperature controller. The thermal dose controller is designed to force the treatment progression at either the actuation or temperature constraints, which is required to minimize the treatment time. The developed controller is applicable to high and low-intensity treatments, such as thermal ablation and thermoradiotherapy. The developed approach is tested using computer simulations for a one-dimensional model of a tumor with constraints on the maximum allowable temperature in the normal tissue and a constrained power output of the ultrasound transducer. The simulation results demonstrate that the proposed approach is effective at delivering the desired thermal dose in a near minimum time without violating constraints on the maximum allowable temperature in healthy tissue, despite significant plant-model mismatch introduced during numerical simulation. The results of in vitro and in vivo validation are reported elsewhere.