Women’s Imaging: Screening Support
By Erik Anderson
Radiology Today
Vol. 25 No. 8 P. 30
AI is poised to address challenges in breast cancer screening.
Radiology departments are increasingly overwhelmed by rising imaging volumes,1 presenting significant challenges to radiologists. Additionally, the industry faces economic pressures such as escalating operational costs and reduced reimbursement rates for mammography services.2 These factors create a challenging environment for radiologists and may contribute to heightened burnout rates.3 Any steps we can take to ease these pressures, especially in the world of mammography, are a win for all involved.
AI is well-positioned to address the challenges facing radiology departments by enhancing workflow efficiency, improving diagnostic accuracy, and reducing the frequency of unnecessary callbacks and false positives. With AI solutions taking the front stage in nearly every facet of health care today, it is essential for radiologists and administrators to meticulously evaluate their AI investments to optimize outcomes and mitigate potential risks.
A Helping Hand
The goal of breast screening is to detect breast cancer as early as possible when it is more likely to be early-stage and potentially easier to treat. AI enhances the process by improving the accuracy of mammography readings with a meticulous search of each slice of the tomosynthesis image set with a high sensitivity.4,5 Mammography AI solutions are designed to help radiologists find more cancers and reduce recalls, leading to more effective screenings.
Beyond these core benefits, AI can help prepare radiologists before they even begin reading by alerting them to what they can expect from each case. This is particularly helpful for workflow management, as radiologists can view AI results on a patient work list and decide which cases to prioritize based on their available time. For example, AI can suggest, based on its own criteria, which cases are likely to be complex vs those that are straightforward. This allows radiologists to prioritize more challenging cases when time allows and handle the simpler cases during busier periods, thus optimizing workflow. There’s also growing interest in using generative AI to assist with report writing, which could further reduce the administrative workload for radiologists, allowing them to focus more on patient care.
Moreover, AI can potentially aid in maintaining compliance with the Mammography Quality Standards Act by detecting potential image quality issues and suggesting improvements. This capability makes mammography AI technology potentially an efficient and effective method for upskilling staff and maintaining compliance.
AI technology is also expanding access to quality mammography. AI-assisted image interpretation brings a new level of standardization to care, leading to more consistent readings and reducing human error. For example, AI’s consistent approach to assessing breast density6 can reduce interreader variability. By providing stable outcomes year after year, AI minimizes discrepancies between radiologists and helps avoid patient confusion.
Notably, two-thirds of mammograms are reviewed by nonfellowship-trained radiologists,7 which adds another layer of complexity to the diagnostic process. Research indicates that fellowship training in breast imaging is associated with greater sensitivity in cancer diagnosis and higher overall accuracy.8 Mammography AI solutions can help narrow the performance gap by helping generalist radiologists achieve improved diagnostic accuracy, closer to that of fellowship-trained breast specialists.4 Thus, the potential use of AI can make quality mammography more accessible to patients in rural or underserved areas who lack access to specialized breast radiologists.
Future Trends
Mammography AI solutions are here to stay and will continue to advance, especially in improving image quality and reducing recalls—two critical goals in breast screening. As AI evolves, it is expected to further ease the administrative burden on radiologists, addressing an important need in today’s health care environment. Looking ahead, here’s what we expect from AI in mammography.
More AI-Driven Personalized Breast Cancer Risk Assessments
AI algorithms have the potential to analyze imaging data for short-term risk assessments, improving current methods that rely only on family history and genetic information. By focusing on image-based evaluations, this approach offers a more immediate and personalized risk assessment, potentially helping to identify high-risk individuals and detect cancer earlier.
More Diverse Training Data Sets
AI will continue to improve by incorporating more diverse training data to avoid bias across different races and ethnicities. A major concern for administrators and radiologists in mammography AI is algorithm bias. Algorithms must be trained on a large, diverse set of images representative of the general population to avoid this bias.
Autonomous AI Review
There is a future path where AI takes an increasingly autonomous role. Currently, laws do not allow fully autonomous or semiautonomous AI in breast cancer screening, presenting challenges for adoption due to legislative barriers.9 AI can make recommendations but, for now, we rely on radiologists to review, confirm, and validate results. While the potential for autonomous AI is promising, the industry must work to thoughtfully address this issue. Even with existing AI-powered technology, the review process still represents a significant burden for radiologists.
Radiologists and administrators should invest in mammography AI tools that support their accessibility, workflow, and quality goals. The future of mammography AI is promising, with exciting developments on the horizon. Collaboration among radiologists, AI developers, and policymakers is key to fully harnessing the technology’s benefits and addressing emerging trends. Together, we can drive advancements in breast cancer detection and enhance patient care.
— Erik Anderson has served as president of breast and skeletal health solutions at Hologic since September 2022.
References
1. Rawson JV, Smetherman D, Rubin E. Short-Term strategies for augmenting the national radiologist workforce authors. AJR Am J Roentgenol. 2024;222(6):e2430920.
2. How are radiology practices impacted by annual changes to the MPFS? AuntMinnie website. https://www.auntminnie.com/practice-management/administration/article/15667525/how-are-radiology-practices-impacted-by-annual-changes-to-the-mpfs. Published May 29, 2024.
3. Fawzy NA, Tahir MJ, Saeed A, et al. Incidence and factors associated with burnout in radiologists: a systematic review. Eur J Radiol Open. 2023;11:100530.
4. US Food and Drug Administration. Genius AI Detection K201019 510(k) summary. https://www.accessdata.fda.gov/cdrh_docs/pdf20/K201019.pdf. Published November 18, 2020.
5. US Food and Drug Administration. Genius AI Detection 2.0 K221449, K230096, 510(k) summary. www.accessdata.fda.gov/cdrh_docs/pdf22/K221449.pdf. Published October 6, 2022.
6. Magni V, Interlenghi M, Cozzi A, et al. Development and validation of an AI-driven mammographic breast density classification tool based on radiologist consensus. Radiol Artif Intell. 2022;4(2):e210199.
7. Lewis RS, Sunshine JH, Bhargavan M. A portrait of breast imaging specialists and of the interpretation of mammography in the United States. AJR Am J Roentgenol. 2006;187(5):W456-W468.
8. Elmore JG, Jackson SL, Abraham L, et al. Variability in interpretive performance at screening mammography and radiologists' characteristics associated with accuracy. Radiology. 2009;253(3):641-651.
9. Responsible steps to implementing AI in breast screening. RSNA website. https://www.rsna.org/news/2024/april/ai-and-breast-screening. Published April 26, 2024.
