Automatic segmentation of cardiac images: texture mapping

Image texture is used to segment left ventricular magnetic resonance (MR) images. The approach is built upon assumptions regarding those probability density functions (PDFs) present within the image. Specifically, an algorithm is constructed which assumes that the data are distributed as fractional Brownian motion (FBM). The FBM is parameterized by H, which quantifies the roughness present in the image texture. When the algorithm is applied to MR data, a bimodal histogram results. This allows automatic classification of the data with resulting LV segmentation.<
>