Effects of Region of Interest on Breast Cancer Detection using CNN and Infrared Imaging

Breast cancer is one of the most common type of cancer that affects woman in the World. It affects millions of women every year killing hundreds of thousands yearly. Early detection of this disease is essential for improving chances of cure and recovery of the patients, thus, one of the most important factors in patient’s treatment is early detection. Mammography, the golden standard for detection this decease is not always 100% effective, which means that it is not always recommend.
In this sense, infrared imaging is a promising technique that might be used as a complementary examination technique in a computer-aided diagnosis system. In this work we used genetic algorithms and convolutional neural networks to classify images from the DMR-IR public database. We analyzed both the entire image and only the breast region. Best results were F1-score of 0.92 for entire images and 0.90 for breast regions.