Deep Hybrid AdaBoost Ensembles for Histopathological Breast Cancer Classification

Breast cancer (BC) is the most common diagnosed cancer type and one of the top leading causes of death in women worldwide. The early diagnosis of this type of cancer is the main driver of high survival rate. This paper aims to use homogenous ensemble learning and transfer learning for binary classification of BC histological images over the four-magnification factor (MF) values of the BreakHis dataset: 40X, 100X, 200X, and 400X. The proposed ensembles are implemented using a hybrid architecture (HA) that combines: (1) three of the most recent deep learning (DL) techniques as feature extractors (FE): DenseNet_201, Inception_V3, and MobileNet_V2, and (2) the boosting method AdaBoost with Decision Tree (DT) as a base learner. The study evaluated and compared: the ensembles designed with the same HA but with different number of trees (50, 100, 150 and 200), the single DT classifiers with the best AdaBoost ensembles and the best AdaBoost ensembles of each FE over each MF. The empirical evaluations used: four classification performance criteria (accuracy, recall, precision and F1-score), 5-fold cross-validation, Scott Knott (SK) statistical test to select the best cluster of the outperforming models, and Borda Count voting system to rank the best performing ones. Results showed the potential of combining DL techniques for FE and AdaBoost boosting method to classify BC in malignant and benign tumors, furthermore the AdaBoost ensemble constructed using 200 trees, DenseNet_201 as FE and MF 200X achieved the best mean accuracy value with 90.36%. © 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.