Using BI-RADS Descriptors and Ensemble Learning for Classifying Masses in Mammograms

Abstract

Abstract. This paper presents an ensemble learning approach for classifying masses in mammograms as malignant or benign by using Breast Image Report and Data System (BI-RADS) descriptors. We first identify the most important BI-RADS descriptors based on the information gain measure. Then we quantize the fine-grained categories of those descriptors into coarse-grained categories. Finally we apply an ensemble of multiple Machine Learning classification algorithms to produce the final classification. Experimental results showed that using the coarse-grained categories achieved equivalent accuracies compared with using the full fine-grained categories, and moreover the ensemble learning method slightly improved the overall classification. Our results indicate that automatic clinical decision systems can be simplified by focusing on coarsegrained BI-RADS categories without losing any accuracy for classifying masses in mammograms.

Diagnosing Breast Masses Using Ica and Non-Image Features

“…III, we have obtained the best performance for the experiment A, with an AUC about 0.8 and an accuracy about 77%, while for the experiment B was obtained an AUC about 0.7 and an accuracy around 69%. As can be seen, our results (for both experiments) are better than those presented in [32] where a single classifier is used. Besides, if we compare with their results for an ensemble of specialist classifiers, our results are better for experiment A, and…”

“…As we stated in the introduction, considering [32,28,29], we can only reliably compare our results with those presented in [32]. Here, the authors selected the cases provided only by one scanner, the Lumysis, for the purpose of data consistency.…”

“…More recently, some other papers describing classification of masses in mammograms (benign and malignant) have been published [32,28,29]. Though these papers were not analyzed in [14], in our view only [32] would have been listed as reproducible. In that study, they used mass shape features and an ensemble system formed by four specialist classifiers based on patient age (young/old) and region of interest (ROI) size (large/small).…”