Accurate, near-real-time temperature measurements are essential for effective monitoring of all types of thermal therapies. Current methods are too slow, subject to substantial errors from motion and fail in fat. We propose: 1) To develop a modification of the PRF technique to allow simultaneous measurement of proton longitudinal relaxation rate, T1. 2) To continue the implementation of model-based linear predictive techniques to utilize the best estimate of tissue thermal properties and linear predictive filtering (e.g., Kalman filter) to obtain the current estimate of the temperature distribution from the prior temperature distribution. Tissue thermal properties will be estimated based upon segmentation from multiple MRI contrasts.