Thermogram Breast Cancer Detection Using Deep Learning Techniques: A review

— In general, using thermal images and applying image processing on them with the help of deep learning models has facilitated the early diagnosis of breast cancer for doctors and has accelerated the treatment process. Since screening has been a challenging and vital issue for a long time, this study has investigated various imaging methods in general and classified each based on their advantages and disadvantages. However, thermal imaging is particularly discussed in this paper. Thermal imaging makes it possible to identify tumors in the early stages by examining the temperature distribution in both breasts. Due to being a non-invasive screening method and not involving any physical touch, injections or the use of special tools during the process, thermal imaging is considered as more preferred among the medical practitioners. The interpretation of thermal images and its classification into categories such as normal and abnormal for cancer diagnosis is carried out by deep learning models such as convolutional neural network (CNN), U-NET network, etc. This article provides a review of recent studies done in the field of breast cancer diagnosis using deep learning models in thermal images. According to the results reported in recent researches, it seems that the combination of U-NET and CNN models enjoys the best result with 99.33% accuracy and 100% sensitivity while the weakest performance goes to Bayesian classification with the accuracy of 71.88% and the sensitivity of 37%.