A COMPARISON OF AUTOREGRESSIVE SPECTRAL ESTIMATION ALGORITHMS AND ORDER DETERMINATION METHODS IN ULTRASONIC TISSUE CHARACTERIZATION

Several autoregressive (AR) methods for spectral estimation were appli ed toward the task of estimating ultrasonic backscatter coefficients f rom small volumes of tissue, Data were acquired from a homogeneous tis sue-mimicking phantom and from a normal human liver in vivo, AR method s performed better at short record lengths than the traditional DFT (d iscrete Fourier Transform) approach. The DPT method consistently under estimated backscatter coefficients at small gate lengths, Burg's algor ithm, the Modified Covariance algorithm, and the Recursive Maximum Lik elihood Estimation algorithm performed comparably. The Yule-Walker alg orithm did not perform as well as these but offered a slight improveme nt over the DFT, Several order determination methods were tested, Thes e included residual variance (RV), final prediction error (FPE), Akaik e information criterion (AIC), and Minimum Description Length (MDL). T he AIC and MDL produced misleading results at higher orders, The RV an d FPE yielded better results. The autoregressive method offers promise for enhanced spatial resolution and accuracy in ultrasonic tissue cha racterization and nondestructive evaluation of materials.