Estimate of Attenuation Coefficient for Ultrasonic Tissue Characterization Using Time-Varying State-Space Model

The images generated from ultrasound pulse-echo signals have long been used to aid clinical diagnosis. Recently, there has been a growing interest in quantitatively determining the acoustic parameters of the tissue as a means of classification and diagnosis. For example, the frequency-dependent attenuation is known to be correlated with different diseases in the liver. In this paper we introduce a new technique for estimating the attenuation coefficient. The effect of attenuation on an interrogating signal with a gaussian-shaped spectrum can be obtained by studying the Wigner distribution of reflected rf data based on a one-dimensional signal model. We show that under the condition that the attenuation varies linearly with frequency, the spectral mean of the reflected signal decreases linearly with time. The estimation algorithm models the pulse-echo signal as the output of a second-order time-varying state-space innovations model driven by white noise. The state coupling matrix A and the output coupling vector C vary with time in a known fashion; moreover, they are also functions of an unknown constant parameter θ. The attenuation coefficient, which is one of the elements of θ, can be estimated directly using a recursive system identification algorithm. The algorithm was verified using both computer-generated synthetic data and in-vivo liver data of known diagnosis. The results show correlation between the estimated parameter and the pathological State Of the tissue.