Desarrollo de un sistema de detección de tumores en mamografías mediante técnicas de inteligencia artificial.

Abstract

The objective of this project is to develop an algorithm with artificial intelligence techniques, such as convolutional neural networks, which were used for the detection and classification of tumors in mammograms and thus to prevent unnecessary invasive intervention in patients with possible diagnoses of breast cancer.

In order to carry out this project, the MIAS database was used, which contains mammography images previously diagnosed by doctors. The segmentation of these images was carried out, thus achieving greater accuracy in the training of the proposed convolutional neural network and the validation of the network was performed with 30% of the segmented images. The percentage of accuracy achieved in the network training was 80% after performing several workouts with different characteristics to improve the percentage of accuracy.