Estimating a parametric model of temperature distribution from an ultrasound image sequence during HIFU therapy

This paper presents a novel method to estimate a parametric model of temperature distribution from a high-intensity focused ultrasound sequence using a Kalman filter approach. The Kaiman filter enables an initial temperature map, derived from (say) the echo-strain method, to not only be smoothed along the heat conduction direction, but to adopt a shape similar to the defined parametric shape of a sensor model. Effects of the model parameters and the covariances of the Kalman filter are investigated. Experimental results on phantom data show that the overall quality of the resulting temperature map is improved using our approach, enabling the extent of tissue "damage" caused by heating to be readily estimated through a visual display