Combinaison de classifieurs pour la reconnaissance des anomalies mammaires

Abstract

Breast cancer is the most common cancer in women, where early detection plays an important role, increasing the chances of a complete cure and in some cases, enabling less intrusive treatment. In recent years, the integration of machine learning (ML) techniques and especially deep learning (DL) have shown promising results in breast cancer classification. The experiments we have carried out in this project combine three transfer learning models (VGG16, Resnet50 and efficient_B0) and five classical supervised learning techniques (MLP, SVM, random forests, decision trees and logistic regression) using haralick features as well as patient age and mammography density. In this work, we used four complete databases. These experiments will enable us to choose the best combination that includes efficient_B0 and SVM. This combination achieves a recognition rate of 96.70%, a precision of 99.6%, a recall of 93.8% and an F1 score of 96.6%.