A model-based displacement outlier removal algorithm for ultrasonic temperature estimation

In using ultrasound echo strain to estimate temperature for monitoring HIFU thermal therapy, block matching algorithms are used to estimate the displacement over two image frames. However, due to the image distortion and artifacts over frames, the displacement will be estimated with outliers which will seriously bias echo strain estimation. This paper introduces a novel model-based algorithm to constrain the displacement estimation. The displacement is initially estimated using correlation analysis and then the outliers on the displacement data will be removed by examining the difference between the data and a parameterised displacement model. Finally the removed outliers will be replaced by the expected values from the model or the interpolation from neighboring normal data. Experimental results on the data form a uniformly heated gel phantom and a HIFU exposed bovine liver showed that the model-based algorithm successfully removed the displacement outliers and was also superior to a median filter which is a common method. In conclusion, the model-based algorithm introduce a displacement model as prior information to remove displacement outliers, which shows great potential to improve the ultrasonic temperature estimation from the ultrasound images distorted by HIFU exposures.