Breast cancer is the most common type of cancer among women and is responsible for one of the largest mortality rates, second only to lung cancer. There are several imaging modalities that can (i.e. ultrasound, tomosynthesis etc), but mammography is currently the most effective tool for early detection of breast cancer.
Despite the efficiency of mammography in breast cancer detection, radiologists still fail to detect between 10% and 30% of malignant breast nodules, mainly due to low contrast in mammographic nodules and human factors, such as fatigue and little experience of the physician. A Computer-Aided Diagnosis (CAD) system for breast cancer detection may contain modules for segmentation and classification of mammographic findings, acting as a Decision Support System (DSS) that helps the medical professional in the interpretation of the exam. With the use of CADs, the efficiency of radiologist's interpretation can be improved in terms of accuracy and consistency in detection/diagnosis, while his/her productivity can be improved by reducing the time required for reading the images.
Machine Learning algorithms are important tools for the detection of malignant tumors in CADs. Shape and texture features have been vastly reported as adequate descriptors for the classification of mammographic findings, however the use of Deep Learning for detection, segmentation and classification of nodules is yet to be fully explored.
The PATREO group is currently developing research projects in biomedical engineering, focusing primarily on the detection of abnormalities in mammographic exams. Our team is using Deep Learning-based methods to preprocess, detect and classify potentially hazardous mammographic findings. This work aims to help physicians in the laborious process of biomedical image analysis.