Evaluation of the spectral fit algorithm as functions of frequency range and /spl Delta//spl kappa/a/sub eff/

Abstract

Considerable effort has been directed at quantifying the properties of the tissue microstructure (i.e., scatterer correlation length) to diagnose disease and monitor treatment. In vivo assessments have had limited success due to frequency-dependent attenuation along the propagation path (i.e., total attenuation) masking the frequency dependence of the scattering from the tissue microstructure. Previously, both total attenuation and scatterer correlation length, given by the effective radius, were solved simultaneously by a two-parameter minimization of the mean squared error between a reference spectrum, modified by the attenuation and scatterer effective radius, and the backscattered waveforms using an algorithm termed the spectral fit algorithm. Herein, the impact of frequency range (largest frequency minus smallest frequency) and deltakaeff (largest kaeff value minus smallest kaeff value; k is wave number and aeff is scatterer effective radius) used by the spectral fit algorithm on estimating the scatterer effective radius, and total attenuation was assessed by computer simulations while excluding frequencies of the backscattered power spectrum dominated by electronic noise. The simulations varied the effective radius of the scatterers (5 microm to 150 microm), the attenuation of the region (0 to 1 dB/cm-MHz), the bandwidth of the source, and the amount of electronic noise added to the radio frequency (rf) waveforms. The center frequency of the source was maintained at 8 MHz. Comparable accuracy and precision of the scatterer effective radius were obtained for all the simulations whenever the same deltakaeff was used to obtain the estimates. A deltakaeff of 1 gave an accuracy and precision of approximately 15% +/- 35%, and a width of 1.5 gave an accuracy and precision of approximately 5% +/- 15% consistently for all of the simulations. Similarly, the accuracy and precision of the total attenuation estimate were improved by increasing the frequency range used by the spectral fit algorithm.