Regresar

Identification of Breast Cancer Through Digital Processing of Mammograms

Abstract:

Breast cancer is one of the main types of cancerous diseases in terms of its mortality rate, although it can normally be cured if the disease is detected early enough. For this reason, it is very important to have computer-assisted diagnosis systems that allow the early detection of breast cancer, especially in underdeveloped countries where women do not have the opportunity to access specialist physicians. Thus, this work proposes a system for the identification of breast cancer on mammograms, whose main objective is to keep the implementation of the system simple, efficient and effective, reducing the need for complex operations or expensive hardware. For this, the current proposal uses digital image processing techniques for the enhancement and segmentation of the regions of interest. Machine learning algorithms (e.g., naive Bayes, artificial neural networks, decision trees, and nearest neighbors) are subsequently used to classify a reduced set of features extracted from the processed mammograms, according to labels that have been validated by specialist clinicians. The evaluation results show that the current proposal is straightforward and, at the same time, precise enough to identify breast cancer on mammograms with accuracy and sensitivity values greater than 99%.