Design and Analysis of Early Detection System of Breast Cancer Based on Mammogram Images using Neural Network

Abstract

The uncontrolled and abnormal growth of the cells leads to the formation of the tumor in our human body. Two types of tumor are formed by uncontrolled cell growth. One is benign and malignant. The malignant is the cancerous cell which can spread to the other body parts very easily. When the cells grow uncontrollably in the breast region that can cause the tumor formation in the breast and finally may happens cancer. In today’s world the second most lethal cancer for human is breast cancer. So early detection and diagnosis is very important to reduce the mortality rate. This paper states that we have designed one new system for early detection of breast cancer from mammogram images based on neural network and we have compares the results with various existing methods. We have divide the work into two parts training and testing phase. In the training phase we have used adaptive median filter, Gaussian Mixture Model (GMM) segmentation and Gray-Level Co-Occurrence Matrix (GLCM) features extraction method. We have used the Mammographic Image Analysis Society (MIAS) dataset to train our system then we have taken some testing mammogram images from some local hospitals and use the same pre-processing method. After that we have used probabilistic neural network (PNN) classifier to classify the tumor whether it is benign or malignant and give the desired output with great accuracy.