AI in Mammo

The $40 Question: What Your Patients Need to Know About AI Mammography Add-Ons

AI in Health care| Women's Imaging
July 23, 2025•12 min read

A Critical Look at the Latest Technology Promising Better Breast Cancer Detection

Your patients may have already encountered it: an offer to pay $40 out-of-pocket for artificial intelligence to review their mammogram. As women's health providers, we need to understand what's behind these offers and whether they truly benefit our patients.

A recent perspective in the New England Journal of Medicine raises important concerns about AI mammography tools and other screening "enhancements" that have entered the market without proven clinical benefits.13 The authors—Benjamin Hyams, Karla Kerlikowske, and Rita Redberg—provide a sobering analysis that every women's center should consider.32

The Current Landscape: AI at the Point of Care

RadNet's Enhanced Breast Cancer Detection (EBCD) service, now in seven states, exemplifies the trend.2 For $40, patients receive an AI analysis of their mammogram using FDA-cleared software called Saige-Dx.1 The marketing promises "the world's most experienced breast specialists" reviewing your scan.2 Similar services are emerging nationwide, with some facilities reporting that 36-40% of offered patients choose to pay for the AI add-on.1

Early data shows women who pay for AI analysis are 21% more likely to have cancer detected.4 This is a relative increase; for context, if a standard screening program finds 5 cancers per 1,000 women, a 21% increase means finding about 1 additional cancer per 1,000 women screened.9 But here's the critical question: Does finding more cancers actually save more lives?

History Repeating: Lessons from CAD and DBT

We've been here before. Computer-Aided Detection (CAD) received FDA approval in 1998 and Medicare coverage in 2002.21 By 2012, CAD was used in 83% of screening mammograms, costing over $400 million annually.22 Yet a landmark study of 625,000+ mammograms found identical cancer detection rates with or without CAD—while increasing false positives and unnecessary biopsies.22

Digital Breast Tomosynthesis (DBT) followed a similar path. FDA-approved in 2011, it's now available in most facilities despite lacking evidence for mortality reduction.26 While DBT may increase cancer detection by 25-29%, the ongoing Tomosynthesis Mammographic Imaging Screening Trial still hasn't answered whether this translates to saved lives.27

The FDA Loophole: Why "Cleared" Doesn't Mean "Clinically Proven"

The NEJM authors highlight a crucial regulatory gap. AI mammography tools like Saige-Dx gain market entry through the FDA's 510(k) pathway, which only requires showing "substantial equivalence" to existing devices—not clinical superiority or patient benefit.15

The FDA's 510(k) summary for Saige-Dx shows its "substantial equivalence" was claimed against another AI mammography tool, not software for wrist fractures.19 However, this highlights a systemic issue with the 510(k) pathway known as "predicate creep." A new device only needs to be similar to a previous one, which was similar to one before it. Over time, this can create a long chain of devices that drift significantly from the original, more rigorously tested technology, all without requiring new clinical trials to prove patient benefit.18 The FDA studies supporting its clearance used image sets with 42% cancer rates—vastly different from the less than 1% rate in actual screening populations.19

What Really Matters: Clinical Endpoints vs. Detection Metrics

The fundamental issue is that increased cancer detection isn't necessarily beneficial. Research shows 11-19% of screen-detected breast cancers represent overdiagnosis—cancers that would never have caused symptoms or death if left undetected.33 This means one in six detected cancers may lead to unnecessary treatment without extending life.34

True clinical benefit requires demonstrating:

  • Reduced breast cancer mortality

  • Fewer interval cancers (cancers appearing between screenings)

  • Decreased rates of advanced cancer at diagnosis

Simply finding more cancers can paradoxically harm patients through overdiagnosis, unnecessary biopsies, and increased anxiety—without saving lives.

The Promise and Peril of AI: What Current Research Shows

Encouragingly, some researchers are asking the right questions. Sweden's MASAI trial, enrolling over 100,000 women, uniquely uses interval cancer rates as its primary endpoint—a true measure of clinical benefit.36 Initial results show AI increased cancer detection by 28% while reducing radiologist workload by 44%, but the critical interval cancer data won't be available until participants have two years of follow-up.37

Until such trials demonstrate mortality benefit, AI mammography remains an expensive experiment conducted at population scale.

Practical Guidance for Your Practice

For your clinical team:

  • Understand that FDA clearance doesn't equal proven clinical benefit.15

  • Know that current AI tools increase detection but lack evidence for improved outcomes.9

  • Be prepared to discuss the concept of overdiagnosis with patients.34

  • Consider that charging patients directly for unproven technology raises ethical concerns.39

For patient conversations:

  • Explain that finding more cancers isn't always better—some detected cancers would never cause harm.34

  • Clarify that no AI mammography tool has proven to save lives or prevent advanced cancer.37

  • Acknowledge the technology shows promise but needs more rigorous study.37

  • Support informed decision-making without pushing patients toward or away from these services.

Key talking points:

  • "This AI technology has been shown to increase the number of cancers found on a mammogram. However, the critical research is still ongoing to determine if this leads to better health outcomes, such as saving more lives or finding dangerous cancers earlier. Right now, we don't have that proof.9"

  • "An important concept in all cancer screening is 'overdiagnosis.' This means we sometimes find slow-growing cancers that, if left alone, would never have caused a problem in a person's lifetime. A major question with any new, more sensitive technology is whether it's finding more of these overdiagnosed cancers, which can lead to unnecessary treatment and anxiety.33"

  • "The FDA has cleared this technology as safe, which means it's similar to other tools already on the market. However, this is different from FDA-approved, which would require proof from clinical trials that it actually improves patient health. That proof doesn't exist yet for AI in mammography.15"

  • "We have seen this pattern before with other new technologies in mammography. They were adopted widely because they found more cancers, but large studies later showed they didn't actually improve survival rates. That's why the medical community is being cautious this time and asking for stronger evidence before recommending widespread use.22"

The Bottom Line

The NEJM authors conclude that "AI reading aids could represent a clinically relevant innovation—or they could cause harm and be an inefficient use of health care resources." Their call for evidence of meaningful clinical benefit before widespread adoption deserves our support.

As women's health providers, we must advocate for technologies that demonstrably improve our patients' lives, not just their scan results. Until AI mammography proves it can reduce mortality or prevent advanced cancer—not merely increase detection—we should view these $40 offers with healthy skepticism.

The authors remind us that our goal isn't finding more cancers; it's saving more lives. That distinction matters profoundly for the women we serve.

Based on: Hyams B, Kerlikowske K, Redberg RF. New Mammography Tools — The Need for Clinically Meaningful Assessment Standards. N Engl J Med. 2025;393:211-213.32

Citations:

1 Hyams B, Kerlikowske K, Redberg RF. New Mammography Tools — The Need for Clinically Meaningful Assessment Standards.

N Engl J Med. 2025;393:211-213. 13

2 RadNet, Inc. (2023, February 8).

RadNet Launches the Enhanced Breast Cancer Detection Program... [Press release]. 2

3 Stempniak, M. (2025, May 15). RadNet confident payers will cover its breast imaging AI solution.

Radiology Business. 1

4 Merchant, T. (2025, May 20). RadNet Tackles Screening Gaps with AI Program.

Axis Imaging News. 9

5 Stocktitan. (2025, February 26).

RadNet and OB/GYN Specialists of the Palm Beaches Enhance Breast Cancer Screening... 11

6 RadNet, Inc. (2025, February 26).

RadNet and OB/GYN Specialists of the Palm Beaches Enhance Breast Cancer Screening... [Press release]. 12

7 The Acorn Staff. (2025, July 18). Health group turns to AI breast cancer detection.

The Acorn. 4

8 Lehman, C. D., et al. (2015). Diagnostic Accuracy of Digital Screening Mammography With and Without Computer-Aided Detection.

JAMA Internal Medicine, 175(11), 1828–1837. 22

9 Lehman, C. (2015, September 28). For mammography, radiologists do better without computer help.

KPWHRI. 24

10 U.S. Food and Drug Administration. (1998).

Summary of Safety and Effectiveness Data: R2 Technologies (P970058). 42

11 Centers for Medicare & Medicaid Services. (2002, October).

Medicare Claims Processing Manual, Chapter 4. 30

12 Rao, V. M., et al. (2018). New Frontiers: An Update on Computer-Aided Diagnosis for Breast Imaging in the Age of Artificial Intelligence.

American Journal of Roentgenology, 212(2), 283-292. 21

13 Lehman, C. D., et al. (2015). Diagnostic Accuracy of Digital Screening Mammography With and Without Computer-Aided Detection.

JAMA Internal Medicine. 22

14 Rao, V. M., et al. (2012). The Cost of Breast Cancer Screening in the Medicare Population.

JAMA Internal Medicine. 22

15 Miglioretti, D. L. (2015, September 28). Computer-aided detection does not improve breast cancer screening.

KPWHRI. 23

16 Hologic, Inc. (2011, February 11).

Hologic Receives FDA Approval for First 3-D Digital Mammography (Breast Tomosynthesis) System. [Press release]. 26

17 McDonald, E. S., et al. (2021). Digital Breast Tomosynthesis: Update on Technology, Evidence, and Clinical Practice.

Radiographics, 41(4), 977–990. 27

18 American College of Radiology. (n.d.).

Tomosynthesis Mammographic Imaging Screening Trial (TMIST). 29

19 U.S. Food and Drug Administration. (2022, May 2).

510(k) Summary for Saige-Dx (K220105). 19

20 Kadakia, K. T., et al. (2023). Medical Devices Cleared on Faulty Predicates More Likely to be Recalled.

TCTMD. 18

21 U.S. Food and Drug Administration. (2021).

510(k) Premarket Notification. 15

22 Ryser, M. D., et al. (2022, February 28). Study Estimates One in Seven U.S. Breast Cancers May Be Over-Diagnosed.

Duke Health. 34

23 Flemban, A. F. (2023). Overdiagnosis in Screening Is the Increase in Breast Cancer Incidence Rates a Cause for Concern? A Systematic Review and Meta-Analysis.

Journal of Personalized Medicine, 13(3), 526. 33

24 Hendrick, R. E. (2024). Overdiagnosis in Breast Cancer Screening: A Critical Review.

Journal of Breast Imaging, 6(4), 338-345. 35

25 Lång, K., et al. (2023). Artificial intelligence–supported screen reading versus standard double reading in the Mammography Screening with Artificial Intelligence trial (MASAI): a clinical safety analysis...

The Lancet Oncology, 24(8), 936-944. 36

26 Lång, K. (2023, August 1). AI-supported mammography screening is found to be safe.

Lund University. 37

27 The Imaging Wire. (2024, March 3).

MASAI Gets Even Better at ECR 2024. 38

28 Schenker, Y., et al. (2014). The ethics of advertising for health care services.

The American Journal of Bioethics, 14(1), 34-43. 40

29 American Medical Association. (n.d.).

Direct-to-Consumer Advertisements for Prescription Drugs. Code of Medical Ethics. 39

30 Caperay. (2022, March 4).

Learning from Past Mistakes. 41

31 RadNet, Inc. (2022, May 16).

RadNet Artificial Intelligence Subsidiaries, DeepHealth and Quantib, Obtain FDA Clearance... [Press release]. 7

Works cited

  1. RadNet confident payers will cover its breast imaging AI solution by ..., accessed July 23, 2025, https://radiologybusiness.com/topics/artificial-intelligence/radnet-confident-payers-will-cover-its-breast-imaging-ai-solution-2026

  2. RadNet Launches EBCD Program Powered by AI - Enhanced Breast Cancer Detection, accessed July 23, 2025, https://myebcdmammo.com/news/radnet-launches-the-enhanced-breast-cancer-detection-program

  3. Radiology Imaging Associates Launches the Enhanced Breast Cancer Detection Program, accessed July 23, 2025, https://www.radnet.com/about-radnet/news/ebcd-florida

  4. Health group turns to AI breast cancer detection - The Acorn, accessed July 23, 2025, https://www.theacorn.com/articles/health-group-turns-to-ai-breast-cancer-detection/

  5. Mammography with EBCD | RadNet Los Angeles, accessed July 23, 2025, https://www.radnet.com/los-angeles/ebcd

  6. Enhanced Breast Cancer Detection | AI Mammography - RadNet, accessed July 23, 2025, https://www.radnet.com/medical-imaging-manhattan/enhanced-breast-cancer-detection-ai-mammography

  7. Two RadNet Subsidiaries Receive FDA Clearance for AI Algorithms | Applied Radiology, accessed July 23, 2025, https://appliedradiology.com/articles/two-radnet-subsidiaries-receive-fda-clearance-for-ai-algorithms

  8. Enhanced Breast Cancer Detection (EBCD) - Breastlink, accessed July 23, 2025, https://breastlink.com/services/enhanced-breast-cancer-detection

  9. RadNet Tackles Screening Gaps with AI Program - Axis Imaging News, accessed July 23, 2025, https://axisimagingnews.com/market-trends/cloud-computing/machine-learning-ai/radnet-tackles-screening-gaps-with-ai-program

  10. Enhanced Breast Cancer Detection | AI Mammography, accessed July 23, 2025, https://myebcdmammo.com/

  11. RadNet and OB/GYN Specialists of the Palm Beaches Enhance Breast Cancer Screening with State-of-the-Art Imaging Services - Stock Titan, accessed July 23, 2025, https://www.stocktitan.net/news/RDNT/rad-net-and-ob-gyn-specialists-of-the-palm-beaches-enhance-breast-t06utr68jhc1.html

  12. RadNet and OB/GYN Specialists of the Palm Beaches Enhance Breast Cancer Screening with State-of-the-Art Imaging Services, accessed July 23, 2025, https://www.radnet.com/about-radnet/news/radnet-and-ob-gyn-specialists-of-the-palm-beaches-enhance-breast-cancer-screening-with-state-of-the-art-imaging-services

  13. New Mammography Tools — The Need for Clinically Meaningful ..., accessed July 23, 2025, https://ramaonhealthcare.com/new-mammography-tools-the-need-for-clinically-meaningful-assessment-standards/

  14. Artificial Intelligence Behind Your Mammogram | Florida - RadNet, accessed July 23, 2025, https://www.radnet.com/radiology-imaging-associates/artificial-intelligence-behind-your-mammogram-florida

  15. Modernizing Medical Device Regulation: Challenges and Opportunities for the 510(k) Clearance Process - PubMed, accessed July 23, 2025, https://pubmed.ncbi.nlm.nih.gov/39374526/

  16. FDA 510(k) Explained: A Basic Guide to Premarket Notification, accessed July 23, 2025, https://www.thefdagroup.com/blog/510k-explained

  17. FDA De Novo vs 510(k) Submission: Which Pathway is Right for Your Medical Devices/IVDs? | Operon Strategist, accessed July 23, 2025, https://operonstrategist.com/fda-de-novo-vs-510k/

  18. Medical Devices Cleared on Faulty Predicates More Likely to be Recalled - TCTMD.com, accessed July 23, 2025, https://www.tctmd.com/news/medical-devices-cleared-faulty-predicates-more-likely-be-recalled

  19. DeepHealth, Inc. B. Nathan Hunt VP, Quality ... - accessdata.fda.gov, accessed July 23, 2025, https://www.accessdata.fda.gov/cdrh_docs/pdf22/K220105.pdf

  20. Diagnostic Accuracy of Digital Screening Mammography With and Without Computer-Aided Detection - ResearchGate, accessed July 23, 2025, https://www.researchgate.net/publication/282306237_Diagnostic_Accuracy_of_Digital_Screening_Mammography_With_and_Without_Computer-Aided_Detection

  21. New Frontiers: An Update on Computer-Aided Diagnosis for Breast Imaging in the Age of Artificial Intelligence | AJR - American Journal of Roentgenology, accessed July 23, 2025, https://ajronline.org/doi/10.2214/AJR.18.20392

  22. Diagnostic Accuracy of Digital Screening Mammography with and ..., accessed July 23, 2025, https://pmc.ncbi.nlm.nih.gov/articles/PMC4836172/

  23. For mammography, radiologists do better without computer help | KPWHRI, accessed July 23, 2025, https://kpwashingtonresearch.org/index.php/news-and-events/blog/2015/sept/news-release-computer-aided-detection-does-not-improve-breast-cancer-screening

  24. Computer-aided detection does not improve breast cancer screening | KPWHRI, accessed July 23, 2025, https://kpwashingtonresearch.org/index.php/news-and-events/recent-news/news-2015/computer-aided-detection-does-not-improve-breast-cancer-screening

  25. Basics of digital breast tomosynthesis | Applied Radiology, accessed July 23, 2025, https://appliedradiology.com/articles/basics-of-digital-breast-tomosynthesis

  26. Hologic Receives FDA Approval for First 3-D Digital Mammography (Breast Tomosynthesis) System, accessed July 23, 2025, https://investors.hologic.com/press-releases/press-release-details/2011/Hologic-Receives-FDA-Approval-for-First-3-D-Digital-Mammography-Breast-Tomosynthesis-System/default.aspx

  27. Digital Breast Tomosynthesis:Update on Technology, Evidence, and Clinical Practice - PMC, accessed July 23, 2025, https://pmc.ncbi.nlm.nih.gov/articles/PMC8170874/

  28. Tomosynthesis Mammographic Imaging Screening Trial (TMIST) - Women's Health Research Institute, accessed July 23, 2025, https://whri.org/tomosynthesis-mammographic-imaging-screening-trial-tmist/

  29. Tomosynthesis Mammographic Imaging Screening Trial (TMIST), accessed July 23, 2025, https://www.acr.org/Clinical-Resources/Publications-and-Research/Clinical-Research/TMIST-Breast-Screening-Trial

  30. Medicare - CMS, accessed July 23, 2025, https://www.cms.gov/Regulations-and-Guidance/Guidance/Transmittals/downloads/R1775B3.pdf

  31. Estimating Breast Cancer Overdiagnosis after Screening Mammography among Older Women in the US - PMC, accessed July 23, 2025, https://pmc.ncbi.nlm.nih.gov/articles/PMC10623662/

  32. New Mammography Tools - The Need for Clinically Meaningful ..., accessed July 23, 2025, https://pubmed.ncbi.nlm.nih.gov/40658666/

  33. Overdiagnosis Due to Screening Mammography for Breast Cancer among Women Aged 40 Years and Over: A Systematic Review and Meta-Analysis, accessed July 23, 2025, https://pmc.ncbi.nlm.nih.gov/articles/PMC10051653/

  34. Study Estimates One in Seven U.S. Breast Cancers May Be Over-Diagnosed | Duke Health, accessed July 23, 2025, https://corporate.dukehealth.org/news/study-estimates-one-seven-us-breast-cancers-may-be-over-diagnosed

  35. USPSTF Recommendations and Overdiagnosis | Journal of Breast Imaging | Oxford Academic, accessed July 23, 2025, https://academic.oup.com/jbi/article/6/4/338/7691827

  36. What is the MASAI (Mammography Screening with Artificial Intelligence trial) study?, accessed July 23, 2025, https://cancer-rose.fr/en/2025/02/13/what-is-the-masai-mammography-screening-with-artificial-intelligence-trial-study/

  37. AI-supported mammography screening is found to be safe | Lund ..., accessed July 23, 2025, https://www.lunduniversity.lu.se/article/ai-supported-mammography-screening-found-be-safe

  38. MASAI Gets Even Better at ECR 2024 - The Imaging Wire, accessed July 23, 2025, https://theimagingwire.com/2024/03/03/final-results-from-the-masai-study/

  39. Direct-to-Consumer Advertisements of Prescription Drugs - Code of Medical Ethics, accessed July 23, 2025, https://code-medical-ethics.ama-assn.org/ethics-opinions/direct-consumer-advertisements-prescription-drugs

  40. Off-Label Use vs Off-Label Marketing of Drugs: Part 1: Off-Label Use—Patient Harms and Prescriber Responsibilities | JACC, accessed July 23, 2025, https://www.jacc.org/doi/10.1016/j.jacbts.2022.12.011

  41. Learning from Past Mistakes - CapeRay blog, accessed July 23, 2025, https://www.caperay.com/blog/learning-from-past-mistakes/

  42. Computer-Aided Detection of Breast Cancer – Have All Bases Been Covered? - PMC, accessed July 23, 2025, https://pmc.ncbi.nlm.nih.gov/articles/PMC3085409/

Artificial intelligence-supported screen reading versus standard double reading in the Mammography Screening with Artificial Intelligence trial (MASAI): a clinical safety analysis of a randomised, controlled, non-inferiority, single-blinded, screening accuracy study - PubMed, accessed July 23, 2025, https://pubmed.ncbi.nlm.nih.gov/37541274/

Back to Blog