MR Image Segmentation: An Algorithm Based on Competitive Hopfield Neural Network and Curve Propagation

摘要

Although Magnetic Resonance Imaging (MRI) can provide high spatial resolution images, the area gray level inhomogenization , weak boundary and artifact often can be found in MR images. So, the MR images segmentation using the gradient-based methods are poor in quality and in efficiency. An algorithm, based on the Competitive Hopfield Neural Network (CHNN) and the curve propagation, is proposed for MR images segmentation in this paper. The algorithm is composed of two phases. In first phase, a CHNN is used to classify the image objects, and to make gray level homogenization and to recognize weak boundaries in objects. In second phase, based the classified results, the level set velocity function is created and the object boundaries are extracted with the curve propagation algorithm of the narrowband-based level set. The test results are promising and encouraging.