Breast detection in X-ray mastography with machine learning techniques

Resumen

Breast cancer is one of the leading causes of death in women in the world. An early detection is crucial to be able to perform an appropriate treatment and thus reduce mortality. An important tool for early detection is mastography, which is a technique to obtain images through X-rays. Through this technique it is possible to detect abnormalities in the breast tissue, thus allowing to detect the disease at an early stage. On the other hand, machine learning algorithms have provided promising results in automatic detection. In this work we combine both X-ray mastography and machine learning algorithms to perform breast detection. The YOLO neural network was used, which is a convolutional network that performs the detection in a single stage. For the training of the network, 216 mastographies from the dataset of breast mammography images with masses were used. Experimental results in position-independent breast detection showed an average confidence of 0.983, and a standard deviation of 0.024 in a sample of 200 images. This result allows to propose YOLO as an accurate preprocessing method for more sophisticated neural networks designed to solve breast cancer related tasks.