Speckle Based Anisotropic Diffusion Filter for Ultrasound Images

Abstract

Imaging of Ultrasound (US) presents significant challenges in visual medical inspection and creation of automated speckle-based analytical approaches that adversely influence tissue boundary detection and the efficacy of automatic segmentation techniques. A number of filtering strategies are usually used as a pre-processing phase before automatic review or visual inspection methods to minimize the impact of speckle. Many state of the art filters seek to decrease the speckle effect without recognizing its significance to tissue structure classification. This loss of expertise is further magnified due to the iterative process of some speckle filters, e.g. diffusion filters, which tend to produce over filtering during the diffusion period due to a progressive shortage of critical details for diagnostic reason. In this one we suggest a filter of an anisotropic diffusion that contains probabilistic-driven memory of probabilistic-driven scheme which can solve problem of over filtering by pursuing philosophy of a selective tissue. In general, we can design formula for the function of memory as a diffusion differential equation for the tensor of diffusion whose behavior depends on statistics of the tissue, by speeding up the cycle of diffusion in unnecessary regions and by utilizing the effect of memory in places where valuable knowledge must have to be stored in reliable manner. Tests of two photos which are real ultrasound and synthetic photos confirm the usage of the mechanism of probabilistic memory to maintain scientifically appropriate frameworks that the state-of-the-art filters are removing.