Breast cancer identification software with CNN
Pereira Carrillo, Jackelin Pamela · 2021
Breast cancer is a serious global health problem to which we are all prone, taking into account the risk factors we are exposed to daily, especially those who work abroad. An incorrect diagnostic could be translated into a bad or inexistent treatment, and in the worst-case flowing into a patient’s death. Nowadays, technological approaches allow us to create and design tools to identify and classify these pathologies using Machine learning methods. Nevertheless, the current neural networks are designed to identify and classify natural objects with different properties than medical images have, causing that the predictions made from them do not have medical validity. For those reasons, this thesis project presents a complete study of the most recently approaches of a breast cancer detector and classifier software, a comparison review between two models of convolutional neural networks, based on modified architectures with our own model; that pretend to adapt to the unique characteristics of medical images based in all the information previously collected, to create a tool that could be useful for radiologist. This work proves the relevance of this technology, its impact into the medical field, and its repercussion and importance of these new tools for the near future of medicine.