Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
Review
. 2025 Jul 1;24(3):279-299.
doi: 10.2463/mrms.rev.2025-0021. Epub 2025 Jun 14.

Application of Machine Learning to Breast MR Imaging

Roberto Lo Gullo  1 , Vivien van Veldhuizen  2   3 , Tina Roa  1 , Panagiotis Kapetas  1   4 , Jonas Teuwen  2   3 , Katja Pinker  1
Affiliations
Review

Application of Machine Learning to Breast MR Imaging

Roberto Lo Gullo et al. Magn Reson Med Sci. .

Abstract

The demand for breast imaging services continues to grow, driven by expanding indications in breast cancer diagnosis and treatment. This increasing demand underscores the potential role of artificial intelligence (AI) to enhance workflow efficiency as well as to further unlock the abundant imaging data to achieve improvements along the breast cancer pathway. Although AI has made significant advancements in mammography and digital breast tomosynthesis, with commercially available computer-aided detection (CAD systems) widely used for breast cancer screening and detection, its adoption in breast MRI has been slower. This lag is primarily attributed to the inherent complexity of breast MRI examinations and also hence the more limited availability of large, well-annotated publicly available breast MRI datasets. Despite these challenges, interest in AI implementation in breast MRI remains strong, fueled by the expanding use and indications for breast MRI. This article explores the implementation of AI in breast MRI across the breast cancer care pathway, highlighting its potential to revolutionize the way we detect and manage breast cancer. By addressing current challenges and examining emerging AI applications, we aim to provide a comprehensive overview of how AI is reshaping breast MRI and improving outcomes for patients.

Keywords: artificial intelligence; breast; deep learning; machine learning; magnetic resonance imaging.

PubMed Disclaimer

Conflict of interest statement

Conflicts of Interest

All authors report no industry support of the project. Katja Pinker serves as a consultant to Bayer, AURA Health Technologies GmbH and Guerbet; has served as a consultant to Neodynamics; received research grants from The Vienna Science and Technology Fund and the National Institutes of Health; and received payment for service on speakers bureaus and for travel/accommodations/meeting expenses from the European Society of Breast Imaging, Bayer and Guerbet. Panagiotis Kapetas received payment for service on speakers bureaus and for travel/accommodations/meeting expenses from the European Society of Breast Imaging, Roentgen Eisenstadt GmbH, and Siemens Healthineers. The other authors of this manuscript declare that they have no conflicts of interest.

Figures

Fig. 1
Fig. 1
Schematic flowchart of the proposed breast DCE-MRI classification system. DCE-MRI, dynamic contrast-enhanced MRI; MIP, maximum intensity projection; T1-wPre, precontrast T1-weighted sequence; TWIST, time-resolved angiography with Interleaved stochastic trajectories; TWISTPre, precontrast TWIST sequence; TWISTN, Nth postcontrast sequence. Reprinted under a Creative Commons (CC BY 4.0) license from Jing et al.
Fig. 2
Fig. 2
DL model triage of screening breast MRI examinations from the external validation data set (A) with BI-RADS and BPE breakdown of examinations classified as “extremely low suspicion” (B) and “possibly suspicious” (C). BI-RADS, Breast Imaging Reporting and Data System; BPE, background parenchymal enhancement; DL, deep learning. Reprinted with permission from Bhowmik et al.
Fig. 3
Fig. 3
ROC curve for the integrated model of clinicopathologic factors, clinical MRI findings, and radiomics features. ROC, receiver operating characteristic. Reprinted with permission under a Creative Commons (CC BY-NC-ND 4.0) license from Lee et al.
Fig. 4
Fig. 4
The workflow for building radiomics analysis-based predictive models. Reprinted under a Creative Commons (CC BY 4.0) license from Peng et al.
See this image and copyright information in PMC

References

    1. Huang G, Liu Z, Maaten LVD, Weinberger KQ. Densely connected convolutional networks. Proceedings of 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2017.
    1. He K, Zhang X, Ren S, Sun J. Deep Residual learning for image recognition. Proceedings of 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2016.
    1. Simonyan K, Zisserman A. Very deep convolutional networks for large-scale image recognition. arXiv preprint arXiv:1409.1556. 2014. 10.48550/arXiv.1409.1556
    1. Ronneberger O, Fischer P, Brox T. U-net: Convolutional networks for biomedical image segmentation. Proceedings of Medical image computing and computer-assisted intervention–MICCAI 2015: 18th international conference, Munich, Germany, October 5-9, 2015, proceedings, part III 18, 2015. 10.1007/978-3-319-24574-4_28 - DOI
    1. Dosovitskiy A, Beyer L, Kolesnikov A, et al. An image is worth 16x16 words: Transformers for image recognition at scale. arXiv:2010.11929. 2020. 10.48550/arXiv.2010.11929 - DOI