An Introduction to Medical Ultrasound and Ultrasonic Scatterer Size Estimation

Anthony Gerig

Until recently, the frequency content of radio frequency signals generated by ultrasonic scatter in tissue has been neglected in favor of simple measures of total reflected intensity. However, such additional information can be utilized to further characterize insonified tissue. My research has been concerned with the development and assessment of methods that can be used to extract parameters related to tissue structure from the power spectra of scattered waves. In particular, the shape of scattered power spectra is determined primarily by scatterer size. By extracting raw data from scanners before processing, and fitting generated power spectral estimates to model curves, sub-resolution tissue scatterer sizes can be estimated. It is envisioned that such estimates, and their corresponding parametric images, would be useful for the diagnosis and monitoring of certain diseases. Not only do such estimates provide numerical information directly related to scatterer properties, but, in some cases, they also may provide additional contrast beyond that of traditional B-mode imaging. My research has focused upon evaluating the accuracy and precision of size estimation techniques, characterizing estimate error in non-ideal scattering environments, and developing methods for increasing estimate precision.

Estimating scatterer size using signals produced by echoed ultrasonic fields inevitably involves preliminary processing devoted to extracting system-dependent factors from spectral estimates to leave quantities that are solely dependent upon the properties of the scattering medium. One purpose of my research has been to investigate the possibility of using a reference phantom to account for these system-dependent factors as an avenue for making the clinical implementation of scatterer size estimation and imaging feasible. My work implemented and verified the accuracy of this method, and investigated its precision through a derived theoretical approximation for size estimate variance that was tested against simulation results. The second purpose has been to characterize, and investigate solutions to, other obstacles that are more intrinsic to scatterer size estimation itself. For example, scatterer size estimates have notoriously low signal-to-noise ratios. Angular compounding (averaging) was investigated extensively as a precision-improving technique that attempts to avoid dramatic sacrifices in resolution. A theoretical expression for the correlation between size estimates generated using data taken from a single region but at different angles of incidence was derived and compared with simulation results. Its utility for compounding optimization and determining relationships between experimental system parameters and measures of performance was explored. The technique was also implemented on a clinical scanner and tested using tissue-mimicking phantoms. Finally, theoretical, simulation and experimental work has been completed on the effects of phase and attenuation aberration, which result from internal variations in sample speed of sound and attenuation, as well as the problem of extracting meaningful size estimates from sample regions populated by scatterers of more than one size.