New developments in artificial intelligence for breast MRI could help unlock efficiencies for radiologists while improving the patient experience for the most common type of cancer.
The power of artificial intelligence (AI) in healthcare has already revealed highly promising advancements that touch on nearly every field of medicine—from identifying patterns that indicate sepsis1 to detecting major events like a collapsed lung.2
In particular, radiology stands to gain a great deal from the potential of AI, thanks to the technology’s ability to identify abnormalities at speeds outpacing the stretches of human potential. Such speed—together with the technology’s high accuracy—can help automate more tedious tasks while enabling clinicians more time to focus on high-level treatment decisions. Those benefits could not only unlock efficiencies, but could also enhance the patient experience thanks to faster scans, easier scheduling, and shorter time-to-hear periods.3
As a result, AI algorithms have been put to use for various imaging modalities, from X-ray to CT scans. AI technologies have also demonstrated their application among magnetic resonance imaging (MRI), and as such, have earned great publicity in the context of prostate cancer, for example.4
Now, with new developments in algorithmic learning for breast cancer, the most common type of cancer5 gets its turn in the spotlight, too.
Algorithmic Learning for Cancer Detection
Recently, research was presented at the 2019 Breast Imaging Symposium that demonstrated the successful use of advanced algorithms on breast MRI, which is traditionally used in cases of particularly dense breast tissue, high breast cancer risk, or after an abnormal mammogram.6
As part of the project, researchers fed MRI images into the “neural network,” a human-like web of algorithms that helps computers learn from errors. It’s a cornerstone of what industry experts call deep learning.
When the technology got something wrong, the images went back through the program until the machine learned its lesson. After repeating that process with some 6,000 image slices, the algorithm essentially reached the point where it could identify the presence of breast tumors in any given MRI image.
Among 277 women studied, the computer yielded an accuracy of 93 percent, a sensitivity of 94 percent, and a specificity of 92 percent. Researchers posited that the outcomes could have been even better had the computer learned from 20,000 slices, as is optimal for deep learning.7
Such positive results could indicate significant improvements on the part of both physicians and patients—in that patients could experience faster review of their case during an otherwise emotional waiting period,8 and that physicians could immediately see MRI slices of interest during imaging workups to speed up reading times.7
Multifaceted Research Combining AI With MRI
Though it was significant, the research presented at the 2019 Breast Imaging Symposium was hardly the first account of AI within the context of breast MRI—and there are other applications of AI and MRI than just detecting abnormalities (though that’s a big part).
For example, in one retrospective review from the American Journal of Roentgenology, authors mapped a history of AI-enabled breast MRI research between 2008 to 2018.9 In that report, which reviewed 67 studies that paired AI with MRI, they found:
- 54 percent of studies investigated AI’s ability to classify lesions among breast MRI.
- 21 percent of studies explored AI’s impact on MRI image processing.
- 13 percent studied AI’s influence on MRI prognostic imaging.
- 12 percent studied the potential role AI-enabled MRI could play in tracking response to neoadjuvant therapy.
Such activity, authors concluded, suggests considerable and varying interest in AI-based research for MRI that will grow in the future.9 As health systems and physicians realize the benefits to outcomes and the bottom line, it may become a consequential piece of MRI—as it could with any other imaging modality.
The Future of AI and MRI
In the future, there may not be a distinction between imaging done with the help of intelligent machines and imaging done without them. Recently, that idea was depicted by a futuristic narrative from GE Healthcare that told the fictional story of Sophie, a woman living in 2030 whose self breast-exam produced abnormal results.
As Sophie begins her care journey, AI not only determines the presence of breast cancer, but it also predicts Sophie’s individual chances of surviving it. That information, made possible by intelligent machines, helps doctors craft the best treatment plan for her—ultimately helping her beat the very disease that took her mother’s life.10
Though the tech involved in Sophie’s story isn’t yet in clinical practice, researchers have made significant headway in studying AI’s potential so that it could be, someday. Additional interest has also picked up among AI-enhanced MRI technologies in other ways, notably with regard to medical image registration and biopsy protocols.11
As those and other efforts continue to rally interest, produce more evidence, and improve AI's applications within the imaging space, it's just a matter of time until AI becomes a ubiquitous part of breast cancer screening—and it could come sooner than you think.
1. The Artificial Intelligence Clinician learns optimal treatment strategies for sepsis in intensive care. Nature. https://www.nature.com/articles/s41591-018-0213-5. Accessed Oct. 4, 2019.
2. AI-embedded X-Ray system could help speed up detection of a collapsed lung. GE Healthcare.
http://newsroom.gehealthcare.com/ai-embedded-x-ray-system-could-help-speed-up-detection-of-a-collapsed-lung/. Accessed Oct. 4, 2019.
3. How radiologists can use artificial intelligence to improve care. American Hospital Association.
https://www.aha.org/news/insights-and-analysis/2019-07-23-how-radiologists-can-use-artificial-intelligence-improve-care. Accessed Oct. 4, 2019.
4. Advancements in MRI for prostate cancer. GE Healthcare. https://www.gehealthcare.com/article/advancements-in-mri-for-prostate-cancer. Accessed Oct. 4, 2019.
5. Common Cancer Types. National Cancer Institute. https://www.cancer.gov/types/common-cancers. Accessed Oct. 4, 2019.
6. Magnetic Resonance Imaging (MRI) - Breast. American College of Radiology and Radiological Society of North America. https://www.radiologyinfo.org/en/info.cfm?pg=breastmr. Accessed Oct. 4, 2019.
7. AI Algorithm Detects Breast Cancer in MR Images. Imaging Technology News. https://www.itnonline.com/article/ai-algorithm-detects-breast-cancer-mr-images. Accessed Oct. 4, 2019.
8. If You’re Called Back After a Mammogram. American Cancer Society. https://www.cancer.org/latest-news/if-youre-called-back-after-a-mammogram.html. Accessed Oct. 4, 2019.
9. Artificial Intelligence for Breast MRI in 2008–2018: A Systematic Mapping Review. American Journal of Roentgenology. https://www.ajronline.org/doi/abs/10.2214/AJR.18.20389. Accessed Oct. 4, 2019.
10. The Future is Now: Imaging in 2030 and Beyond. GE Healthcare. https://www.gehealthcare.com/article/the-future-is-now-imaging-in-2030-and-beyond. Accessed Oct. 4, 2019.
11. How AI will change MR imaging. GE Healthcare. https://www.gehealthcare.com/feature-article/how-ai-will-change-mr-imaging. Accessed Oct. 4, 2019.