Enhanced Hopfield Neural Network with Edge Preserving (EHNN-EP) based severity diagnosis of Ductal Carcinoma in Situ (DCIS)

Abstract

Computer aided mammogram image cancer segmentation is more complex in intensity mapping to predict the normal and infected region. Generally, image thresholding and static feature based clustering deals with the fixed level of intensity mapping to segment it. Since, the pattern structure of given testing image must need to analyse the cancer level. In this paper, the Enhanced Hopfield Neural Network model with Edge Preserving (EHNN-EP) technique is used for segmenting the cancer region from mammogram image which is to enhance the prediction range of image clustering. Initially, the additive noise can be eliminating by median filter which makes the image smoothening and improve the intensity level. This type of enhancing the image leads to provide the edge details of that image. Also, the HNN performs the repeated learning od image feature which improves the image clustering. The performance report of this proposed method of EHNN can be validate by referring the comparison result of traditional state-of-art methods in two different mammogram image databases. The comparison result represents the performance level of EHNN-EP method that achieves the accuracy percentage as 98.56%.