DETECTION AND ANALYSIS OF BREAST CANCER USING CONVOLUTION NEURAL NETWORK FOR MAMMOGRAM IMAGING SYSTEM

Abstract

Breast cancer is one of the most common life-threatening diseases in women, as well as a leading cause of cancer death. Mammography is one of the most effective diagnostic methods for detecting breast cancer. However, breast cancer researchers have advocated for the use of intelligent-based techniques by medical experts and radiologists over the last decade. Deep learning, convolutional neural network (CNN), is providing effective performance in accurately classifying mammograms, which can assist imaging specialists. The CNN model should be trained with a larger number of labelled mammograms to achieve an accurate classification of mammograms. However, it is not always possible to obtain additional labels for mammograms. The primary goal of this experiment is to use CNN with dense layers to perform highly accurate mammogram classification. In our research, we created classification CNN-based models with a single dense layer. In this case, the dense layers serve as the foundation for the CNN model's accurate mammogram classification. This work aims to improve the performance of CNN with more dense layers, including multi-view preprocessed mammograms. We have used CNN with multiple activation functions (like Linear, Sigmoid, SoftMax, Relu etc.) and their combinations to improve the performance of the model and have achieved the maximum accuracy of 85% of the model which uses the activation function in combination of Relu and Sigmoid.