Experimental estimation of effective scatterer diameters from physical phantoms using autoregressive spectral analysis

One of the limitations of tissue characterization using backscatter coefficients (BSCs) is the relatively low spatial resolution of the technique. Although short gate lengths (Δz) can be used in order to improve the axial resolution, this approach may result in the distortion of the estimated BSC curves when using the discrete Fourier transform (DFT) for spectral estimation. Previous experimental reports suggested that autoregressive (AR) methods can improve the quality of average BSC (ABSC) estimation when using low Δz values. In this study, the estimation of effective scatterer diameters (ESDs) from a calibrated physical phantom using AR spectral analysis was experimentally assessed. Backscatter data was obtained from an agar-based phantom with embedded 41± 4 μm glass spheres using single-element focused transducers with 5 MHz, 7.5MHz and 10 MHz nominal center frequencies. BSCs were estimated using both DFT and AR methods from 21 analysis regions containing 21 scan lines and Δz values between 0.5 and 18 wavelengths (λ). The ABSC and ESD values were derived from the BSCs using both the -6 dB and -12 dB bandwidths of each transducer. ESDs were estimated from the BSC curves by using a solid sphere model. AR estimation was performed using Burg's method and the AR order was selected to optimize the fit between the estimated and theoretical BSC curves. The results showed that AR model typically outperfomed DFT at very short gate lengths using any of the three parameters. Nevertheless, the results suggest that improvements in ESD estimation when using AR analysis are dependent on the analysis bandwidth, and may not necessarily correlate with improvements in ABSC estimation.