부산대학교 도서관

High frequency A-scan lines were simulated considering point-like diffusers in an attenuating medium in order to mimick skin tissues. Then, the centroid algorithm coupled to a second order autoregressive model (AR2) was used to estimate local attenuation coefficients. Attenuation coefficients were calculated for several temporal window lengths, for various attenuations, and for frequencies ranging from 20 to 100 MHz. The best results were obtained for 80 MHz and 45 MHz, where the estimated attenuations were very close to the theoretical values. At 80 MHz, the relative error is less than 6% for window sizes greater than 0.1 /spl mu/s except for 4 dB/cm.MHz, where the error is around 10%. At 45 MHz, the relative error is always less than 8% for the same window sizes. The above estimator was then applied to backscattered signals at 35 MHz obtained from 2 mm thick human skin in-vitro. The first results indicate that the estimation of the attenuation of cutaneous tissue from backscattered signals is feasible but that diffraction correction and calibration on a phantom of known attenuation are necessary to obtain sufficient precision.