New Deep Learning-Based Algorithm Can Assess Breast Density

Researchers from the University of Pittsburgh Medical Center (UPMC) (Pittsburgh, PA, USA) have developed a new deep learning-based algorithm for breast density segmentation and estimation that correlated well with BI-RADS density assessments by radiologists and also outperformed an existing state-of-the-art algorithm.

The algorithm used a fully convolutional network, which is a deep learning framework for image segmentation, to segment both the breast and the dense fibroglandular areas on mammographic images. Using the segmented breast and dense areas, the algorithm computed the breast percent density (PD), which is the faction of dense area in a breast. Using full-field digital screening mammograms of 604 women and the validation dataset, the researchers evaluated the performance of the proposed algorithm against the radiologists' BI-RADS density assessments. Specifically, they conducted a correlation analysis between a BI-RADS density assessment of a given breast and its corresponding PD estimate by the proposed algorithm. In order to demonstrate the effectiveness of their algorithm, the researchers also compared the performance of their algorithm against a state-of-the-art algorithm, LIBRA.

The researchers found that the PD estimated by their algorithm correlated well with BI-RADS density ratings by radiologists and also outperformed LIBRA. Their algorithm provided reliable PD estimates for the left and the right breast and showed excellent ability to separate each sub BI-RADS breast density class. The researchers now plan to release the algorithm to the public through a program/resource sharing service such as GitHub.

Related Links:
University of Pittsburgh Medical Center