1364. A Review of Artificial Intelligence in Mammography
Authors* Denotes Presenting Author
Meghan Jairam *;
Brigham & Women's Hospital
Richard Ha;
Columbia University
Objective:
Breast cancer is the most common cancer among women worldwide. Mammography is a commonly used modality in the detection of breast cancer. Over the past decade, convolutional neural networks (CNNs) and other methods of deep learning have been shown to have a greater detection accuracy than traditional computer-aided detection (CAD). CNNs, as a form of deep learning, have shown the greatest success. Given artificial intelligence (AI’)s rapidly growing presence within mammography, in this systematic review, we aim to summarize the latest developments in the field and shed light on where future progress may lie.
Materials and Methods:
A search in MEDLINE/Pubmed was conducted to identify studies performed from 2017 to present. The studies pertained to the use of AI for improvement in diagnostic metrics, interval cancer rate, triaging, and performance of radiologists when evaluating digital mammography (DM) and digital breast tomosynthesis (DBT).
Results:
Compared to conventional CAD, AI-based CAD has shown a reduced false positive rate, increased sensitivity, and area under the curve (AUC) for DM. When used as an assistive tool for breast cancer detection, AI has shown to improve the sensitivity and AUC for radiologists in both DM and DBT. AI systems have also improved the detection of interval cancers. Many of the articles indicated that deep learning can be used in triaging studies by discarding normal mammograms while maintaining a high degree of accuracy. Although in some studies radiologists outperformed the AI system, in others AI was found to be as accurate or sometimes more accurate when used to independently evaluate studies.
Conclusion:
This review details the most recent developments within AI as applied to mammography, including tomosynthesis. We discuss challenges to widespread adoption of AI in breast imaging, including the cost of implementation as well as the ethical and legal implications of informed consent, bias, liability, privacy, and cybersecurity. The breast imaging community can use this article as a resource for the state of AI in mammography.
