Estimation of ultrasound attenuation coefficient using log-spectrum domain processing

Ultrasound attenuation coefficient is an important diagnostic parameter in medical ultrasonography. Furthermore, it is a parameter of a component related to the attenuation process of the space-variant point spread function, which can be used to improve the spatial resolution of ultrasound images through deconvolution. A recently published approach to the estimation of the ultrasound attenuation coefficient from B-scan radiofrequency data is extended and explained in a detail. First, a parametric image of the mean attenuation coefficients between the probe and a given pixel position is computed by applying linear regression to log-spectra of short segments of radiofrequency signals. Three methods of forming the parametric image are presented. As a second step, the local tissue-specific attenuation coefficients are estimated in small regions of the obtained parametric image. The method has been tested on synthetic radiofrequency data and on radiofrequency data recorded from a tissue-mimicking phantom. A fairly good correlation with the known reference values was achieved.