Two-Dimensional Fuzzy Clustering for Ultrasound Image Segmentation

Segmentation of ultrasound images is a challenging task due to the low signal-to-noise ratio of images, which is mainly caused by two factors: speckle noise and intensity inhomogeneities. A novel segmentation algorithm based on the fuzzy logic is proposed for simultaneously reducing speckle noise and compensating for intensity inhomogeneities. A two- dimensional fuzzy C-means (2DFCM) algorithm with spatial constraints provided by an enhanced speckle reducing anisotropic diffusion (SRAD) filter is firstly developed. Then, by assuming a multiplicative model of differential signal attenuation and a log-compressed model of displayed ultrasound images, the objective function of the 2DFCM is modified to construct a homogenized 2DFCM (2DHFCM). Experimental results on both synthetic image and fetal aortic arch image are given and compared with ones obtained by the gradient vector field (GVF) snake. Experimental results show the effectiveness of the proposed algorithm.