Predicting Mammographic Breast Density Assessment Using Artificial Neural Networks

Introduction: Mammographic density is a significant risk factor for breast cancer. Classification of mammographic density based on Breast Imaging Reporting and Data System (BI-RADS) is usually used to describe breast density categories but the visual assessment can have some restrictions in a routine check in the screening mammography centers. The object of this study was to investigate the effectiveness of artificial neural networks in predicting breast density, based on the clinical patient dataset in a University hospital. Material and Methods: In this study, mammographic breast density was assessed for 219 women who underwent digital mammography screening using Volpara software. A model based on the Multi-Layer Perceptron Neural Network was trained to predict patient density by identifying the (dense vs. non-dense) breast density categories. The predictive model applied to the classification was examined by the Receiver operating characteristic (ROC) curve. Results: The results show that the model predicted the breast density of patients with a classification rate of 98.2%. In addition, the area under the curve (AUC) was 0.998, signifying a high level of classification accuracy. Conclusion: The use of artificial neural networks is useful for predicting patients breast density based on clinical mammograms.