| dc.description.abstract | The area of applied computing has been making an increasing and assertive contribution
in many areas of health. Specifically, in oncology, artificial neural networks, a segment of
artificial intelligence, in the last decade, objectively aggregated the prediction and prognosis of malignant neoplasms / cancer. However, in this area, there are still demands to be resolved and disruptive technologies are a great ally to achieve the expected results. In the field of malignant breast cancer, immunohistochemistry is the most practiced due to its accurate profile with regard to patient staging. Staging means evaluating and classifying the degree of tumor dissemination, aiming at an individualized treatment for each patient. Seeking to eliminate the levels of subjectivity involved in the diagnosis of immunohistochemistry and in order to reproduce the daily routine of pathologists, the work proposes the use of two artificial neural networks: a support vector machine and an Mask R-CNN. Two texture algorithms: local binary pattern and haralick were used to extract the resource vector used as input into the artificial neural network SVM and their results compared. The work shows that comparing only the texture algorithms, haralick showed a better accuracy, which was 81% against 80% of LBP. The use of the mask R-CNN model through the TL technique performed much less than expected for the IHQ data set with mAP for training of 0.050 and for testing of 0.049. | en |
