iagnosis of breast cancer using Artificial Neural Network

Resumo

Breast cancer is a common invasive cancer in women, affecting more than 10% of the world's female
population, being one of the main causes of death in the world. The analysis of the biopsy procedure carried out
by a qualified professional, still appears as the most effective method in the diagnosis of cancer in general, if

diagnosed and treated prematurely, the patient has great chances of cure, however the process can be time-
consuming and the diagnosis is not it's always assertive. In this context, automatic breast cancer classification

algorithms appear, using machine learning algorithms in search of an efficient result for the final diagnosis.
However, it is still necessary to establish methodologies with greater precision to guarantee the assertive result.
Here, we use a multilayer artificial neural to classify breast cancer in two stages: malignant and benign. To test
the proposed methodology, a Breast Cancer Wisconsin (Diagnostic) Data Set database was used, where feature
extractions were made from digitalized images of breast masses. Several architectures of the neural network
have been proposed. And a cross-validation was implemented with 10 k-fold. The result showed an average
accuracy of 92%. Thus, with this result, a tool can be made that can expedite the diagnosis of the patient in a
precise way and, consequently, enable an early treatment increasing the chances of cure.