Using a Two-Layer Competitive Hopfield Neural Network for Medical Image Edge Detection

Abstract

In medical applications, detection and outlining of boundaries of organs and tumors in CT and MRI images are prerequisite. In this paper, a specifically designed two-layer Hopfield neural network called competitive Hopfield edge finding neural network(CHEFNN) is presented for finding the edges of CT and MRI images. Diggerent from the conventional 2-D Hopfield neural networks, CHEFNN extends the one layer two-dimensional Hopfield network at the original image plane into a two-layer three-dimensional Hopfield network with edge detection to be implemented on its third dimension. With the extended 3D architecture, the network is capable of taking each pixel's contextual information into pixels' labeling procedure. As CHEFNN takes pixel's contextual information into its consideration, the effect of tiny details or noises will be effectively removed. As a result, the drawback of disconnected francitons can be avoided. Futhermore, due to the incorporation of competitive learning rule to update the neuron states to avoid the trouble of having to satisfy strong constraints, facilitate the network to converge fast. Our experimental results show that the CHEFNN can obtain more appropriate, more continued edge points than Marr-Hildreth's and Laplacian-based methods.