Machine Learning for Breast Cancer Early Detection

Abstract

Breast cancer is the most common cancer among women in the world. It is also among world’s second most occurring cancer in all types of cancer. The project was first thought of because of the increase in cases of breast cancer, if we can detect it as soon as possible, there will be a better chance of it getting cured which is very important in saving lives. Doctors can use deep learning to accurately diagnose the disease. For breast cancer detection, we can do classification on mammogram images. Neural networks such as pretrained network and convolution neural network can be used for detection as one of the popular techniques in deep learning. How to tune or modify a neural network in order to output a better accuracy is what this project will focus on. In the project, Mammograms-MIAS dataset, including 200 normal images and 154 abnormal ones, is used. The efficiency of machine learning for breast cancer detection in mammogram images, through different testings, have been observed by experimental results. In order to observe the performance of the neural network in an all-round view, comparison will be made between different networks, different dataset and different network structure. Fine tuning pretrained network is a popular method by using transfer learning, the result be will be compared with self-trained network and the architecture of the neural network can be modified and optimised in order to achieve higher accuracy. Moreover, pre data processing such as cropping the centre part of the images will expectantly lead to higher accuracy measures.