Early Stage Detection and Classification of Breast Cancer

Abstract

One of the major diseases that affect young to old aged women in recent times is breast cancer. It almost ranks as the first cause for death in women across the world. The survival rate of people suffering with it ranges somewhere between 40% and 60% depending on the development terms of particular countries. Hence, it becomes quite important to be able to diagnose such a disease at a stage as early as possible, so the patient could look out on the available options for treatment. Therefore, in this project, we propose such a breast cancer detection system which predicts the nature of the cancer, either benign or malignant by processing the mammographic image of the patient. The model basically uses a range of digital image processing techniques and also algorithms of ML in the process to output the prediction. It is trained using the MIAS breast cancer dataset. The input image is first resized, gray-scaled, and a gaussian filter is applied on it to remove background noises. It is then segmented and fed to the neural network, which gives the output prediction as an integer value (each value corresponding to a predicted class). The project also has a second stage where the severity of the cancer is also detected by taking input of other detailed attributes of the mammogram.