Using speckle statistics to improve attenuation estimates for cervical assessment

Abstract

Quantitative ultrasound parameters like attenuation can be used to observe microchanges in the cervix. To give a better estimate of attenuation we can use speckle properties to classify which attenuation estimates are valid and conform to the theory. For fully developed and only one type of scatterer, Rayleigh distribution models the signal envelope. But in tissues as the number of scatterer type increases and the speckle becomes unresolved Rayleigh model fails. Gamma distribution has been empirically shown to be the best fit among all the distributions. Since more than one scatterer type is present for our work we used a mixture of gamma distributions. EM algorithm was used to find the parameters of the mixture and on basis of that the tissue types were segmented from each other based on the criteria of different scattering properties. Attenuation estimates were then calculated for tissues of the same scattering type only. 67 Women underwent Transvaginal scan and the attenuation estimates were calculated for them after segregation of tissues on scattering basis. Attenuation was seen to decrease as the time of delivery came closer.