An integrated framework for classifying mammograms according to BIRADS scale and breast tissue density.
Ioannis N. Tzortzis, Stavros Sykiotis, Ioannis Rallis, Nikolaos D. Doulamis · 2023
In this work, we present an integrated framework for classifying digital or digitized mammograms according to either BIRADS scale or breast density score. We propose unified data preparation procedure, a common training process that includes a Convolutional Neural Network approach and a simple inference tool aiming to ensure the robustness of out model and provide a more generalized solution. We utilize not only a private dataset for training and validation, but also the open dataset INBreast for the testing procedure. According to our results, our solution achieves; a) accuracy score 87% and F1 score 82% on Density classification, and b) accuracy score 85% and F1 score 80% on BIRADS classification.