Machine Learning in the identification of breast tumor images

Introduction: Technology has become part of our daily lives, such as Machine Learning, which uses algorithms, which are rules sequences or instructions that lead step by step to solving a problem. This artificial intelligence can organize and analyze data, detect patterns and make machines learn from them and generate smart solutions. This image recognition technology can be used to identify breast cancer, cervical cancer, prostate cancer, among other types of tumors. Objective: In the present study we proceed with the use of machine learning in the identification of normal and tumor breast tissues. Methods: The validation analyzes of Machine Learning algorithms, using the Python programming language along with Tensorflow, were performed in histopathological images of ductal carcinoma in situ (n=10) from Pathological Atlas online and Padre Albino University Center (UNIFIPA) of Catanduva. Normal breast samples (n=10) were obtained in histological Atlas. Results: The results obtained indicate that the training and accuracy validation tests, performed in the tumor and normal images, have low accuracy to find the patterns related to the ductal in situ carcinoma. In addition, it is important to note that, despite the promising results, the tests of this training scenario show that more training is needed to reduce the loss of accuracy. Conclusion: The set of our data winds that the use of machine learning can contribute to the medical field, improving the accuracy of pathological diagnoses. Validation of the algorithms presented underfitting standards, where it was not able to learn enough about the data. These results stimulate the continuity of studies increasing the number of samples.