The increase of nuclear size is observed in biopsies of patients with benign and malignant diagnosis, as well as, a change in the texture of the nucleus. In this article, an analysis of the variables radius mean (mean of distances from center to points on the perimeter) and texture (standard deviation of gray-scale values) is made using the unsupervised learning algorithm, Ex-pectation-Maximization (EM). Since we observe that those variables have a similar behavior to the mixture of normals in two components. Such algorithm is able to discriminate the data into two groups (malignant and benign). Said model projects a classi cation with a high percentage of coincidence with the observed data.

Maximum likelihood estimators, breast cancer, EM algorithm, Gaussian mixture model