MULTIPARAMETRIC HUMAN LIVER FIBROSIS IDENTIFICATION FROM ULTRASOUND SIGNALS

摘要

Classical acoustic parameters (wavespeed, attenuation and backscattering coefficients and mean scatterer spacing) have been used in the characterisation of biological tissue pathologies. Liver fibrosis is a common disease, classified according to the Metavir Scale (from F0: normal, to F4: cirrhosis). The present work evaluates fibrosis discrimination by combining the mentioned parameters and applying a discriminant analysis on ultrasonic backscattering signals from 20 in vitro human liver samples. The highest correct classification percentages (85-90%) happened for parameters combination in groups of there. The limits and potentiality of the study are also discussed. A greater number of samples and non-linear classification methods are presently being pursued.