CMANTEC NEURAL NETWORK ALGORITHM BASED CLASSIFICATION OF MAMMOGRAMS USING SLBP
J Jeya Caleb M Kannan
A new method for the prediction of breast cancer is presented in this research work which uses Constructive Neural Networks (CNN). As far as the traditional artificial neural networks are concerned, the main problem is in identifying suitable and compact neural network architecture. All previous methods used more layers in the hidden section and also used inter nodes which increased the memory and computational time. The fundamental method i.e. trial and error method still plays a vital role which confronts with two main problems of over-fitting or under-fitting. CNN generates the topology in the training phase by introducing neurons. Here, mammogram images are used from which the texture features are extracted using SLBP (Support Local Binary Pattern). These texture features are then analyzed with predefined learning obtained from neural networks. The best possible combinations of features which can become cancer affected are most likely to be identified as cancer affected cell. These parameters are then used to train the network with the help of CMantec (Competitive Majority Network Trained by Error Correction) algorithm and are thereby used to identify the breast cancer at an improved speed. The effective use of CMantec algorithm brings out efficiently the training process and improves generalization capability for the efficient extraction of texture features of mammogram images
This article is written in Adobe PDF format ( .pdf file ).To view this article you need to download the file. Please rightclick on the link below and then select "Save
target as" to download the file to your harddrive.