Ultrasound attenuation estimation using the LMSE filters and the median filter

The ultrasound attenuation coefficient in known to increase with frequency in soft biological tissue within the diagnostic frequency range. The frequency dependent attenuation coefficient is an important parameter in clinical tissue characterization, especially in liver, and in ultrasonic quantitative B-scan imaging. But, because of the random spatial variation in the backscattered signal amplitude, the estimator variance becomes very large, making it difficult to estimate the attenuation slope from the backscattered ultrasound signal. In this paper we suggest a new ultrasound attenuation estimator based on the calibrated log spectral difference. The proposed estimation process consists of Wiener filtering, Kalman filtering, and median filtering in this order. The simulation results and experimental results with tissue equivalent (TE) phantoms show that combination of the above filterings yields a considerably reduced estimator variance compared with the conventional estimators.