Analysis of Neoadjuvant Treatment Response in Breast Cancer Using Deep Networks

• Kleber S. Pires UFABC
• Francisco A. Zampirolli UFABC
• Fabio A. Marchi ICESP

Resumo

Breast cancer is the most common type of cancer among women, and it has the highest death rate among all types of cancer. Pre-surgical (neoadjuvant) treatment can improve the patient's prognosis, but it is impossible to predict whether the patient will respond. Previous studies were conducted to find characteristics capable of associating neoadjuvant treatment with the patient's response, some using conventional radiomics and others using convolutional networks, mostly using private databases. In this work, a deep learning model was proposed using images and clinical data from a public database (Duke Breast Cancer MRI) capable of extracting characteristics of breast MRI images and associating the attributes with prognosis. The 300 selected patients were divided between training, validation using cross-validation, and testing. Using quantitative analysis of results generated from the trained model, it was concluded that the proposed model can classify patients that achieved a complete response to neoadjuvant treatment. The results demonstrated superior accuracy when compared to the study by Cain et al. in the same database, with a mean AUC (area under the curve) of 0.70 to 0.82 and a mean accuracy of 70% for testing. The proposed model obtained competitive results compared to the literature in public databases, but a further study should be conducted to validate the method in another database.