Breast cancer risk in high risk women is significantly increased for those with high background parenchymal enhancement (BPE) compared to low BPE, yet there is only fair to moderate inter-rater agreement for BPE assessment among radiologists, limiting the use of BPE as a marker of cancer risk. We developed a deep learning algorithm that classifies BPE with high accuracy. The algorithm works best when sub-MIPs, not MIPs, are used as network inputs. The algorithm has potential to autopopulate breast MRI reports in our breast imaging clinic, and ultimately to standardize BPE as a marker of breast cancer risk.
This abstract and the presentation materials are available to members only; a login is required.