Early Stage Breast Cancer Prediction Using Image Processing

Breast cancer is the most frequent cause of death in women, being the second leading cause of cancer deaths worldwide. Early detection is a good remedy hence we have devised a Computer Aided Detection (CAD) method to detect not only cancer but also the type and stage of cancer for early and immediate treatments and classify them into four classes namely benign mass, benign calcification, malignant mass and malignant calcification. Our CAD system is based on Convolutional Neural Network (CNN) for image classification with ResN etSO Architecture. The data is obtained from an open-source database TCIA (The Cancer Imaging Archive) and the dataset was DDSM (CBIS-DDSM). Training on a large dataset gives high accuracy but due to limited image data we used data augmentation techniques to create around 8,000 images from 2,400 images. The image preprocessing steps include DICOM to PNG conversion, resizing, and pectoral muscle removal. The accuracy is 81 % on the CBIS-DDSM dataset.