For decades, a woman’s breast cancer risk score has come from a clipboard: age, family history, age at first period, maybe a biopsy history, fed into models like Gail or Tyrer-Cuzick. Clairity Breast AI risk prediction changes the input. It reads the mammogram the patient already gets and returns a 5-year risk estimate with no questionnaire. Coverage of AI moving breast care “from scan to treatment plan” has put new attention on Clairity as it rolls out to imaging centers. If we run, invest in, or sell into a radiology practice, this is where risk assessment stops being a form and becomes a product line.
What’s actually new with Clairity Breast AI risk prediction
Clairity Breast received FDA De Novo authorization in mid-2025. It was the first AI platform authorized to predict a woman’s five-year breast cancer risk from a standard 2D screening mammogram. Dr. Connie Lehman co-founded the company. Her academic work at Massachusetts General Hospital showed that deep learning models can find risk signals in mammograms that radiologists cannot see and questionnaires never capture. The De Novo route matters because Clairity had no predicate device to compare against. It created a new regulatory category, and later competitors can use it as their predicate.
Since authorization, the change is deployment. Clairity is moving from validation studies into commercial rollout at imaging centers and health systems. The pitch is simple. The software analyzes the screening exam in the background and produces a risk score. That score can trigger follow-up pathways: supplemental MRI, more frequent screening, genetic counseling referral, or a high-risk clinic visit. It needs no extra imaging, no extra patient time, and no front-desk questionnaire that half the patients fill out incompletely.
This is a different job from the AI most practices already know. Tools like Lunit INSIGHT MMG and iCAD ProFound AI Detection help radiologists find cancer that is already there. Clairity estimates the chance that cancer will develop over the next five years. One tool works on today’s read. The other works on the next five years of the patient relationship.
Why it matters
- Risk assessment becomes universal, not selective. Questionnaire-based models depend heavily on family history, which many women don’t know or don’t report. An AI mammogram breast cancer risk score runs on every screening exam, so it identifies more high-risk women, including women with no family history at all.
- Supplemental MRI referrals get a new trigger. Guidelines tie supplemental MRI eligibility to lifetime risk thresholds, and the FDA’s national breast density notification rule took effect in September 2024. As a result, more patients ask about extra screening. Image-based risk gives practices a defensible, data-driven way to decide who gets dense breast supplemental screening and who doesn’t.
- Equity becomes a selling point. Developers built traditional risk models mostly on data from white women, and those models have underperformed in Black, Hispanic, and Asian populations. Lehman’s research group has made performance across diverse populations a core claim. Health systems with equity mandates will ask about this.
- New revenue, unsettled reimbursement. No established, broadly reimbursed CPT code exists yet for AI-derived 5-year risk scoring. Early adopters will lean on self-pay add-ons, bundled “enhanced screening” packages, value-based contracts, and the downstream MRI and genetics volume the score generates. Billing teams need a plan before launch, not after.
- Workflow and liability shift. Once a practice generates a high-risk score, it owns the follow-up. A risk score with no referral pathway is a liability, not a feature.
- Vendor stack consolidation. Practices already paying for detection AI, density assessment, and reporting software will push vendors to bundle. Expect pressure on standalone tools and more partnership announcements.
How to use it today
The vendor and the PACS team handle software integration. Our work is everything around it: evaluation, pathways, patient messaging, and billing. Here’s a practical rollout sequence for an imaging center or a company selling into one.
-
Verify the regulatory status yourself. Don’t rely on marketing decks. Search the FDA De Novo database for the device and read the decision summary, especially the indications for use, the population studied, and the limitations. Use an LLM to extract the key points from the PDF:
You are a regulatory analyst. Summarize this FDA De Novo decision summary for a radiology practice owner. Extract: 1. Exact indications for use (verbatim) 2. Patient population and exclusions (age, prior cancer, implants, imaging type) 3. Validation dataset size, sites, and demographic breakdown 4. Reported performance metrics (AUC, calibration, sensitivity at thresholds) 5. Stated limitations and special controls 6. Anything that restricts how the risk score may be used clinically or marketed Flag any claim a sales rep might make that is NOT supported by this document. -
Run a structured vendor evaluation. When comparing AI radiology tools for imaging centers, send every vendor the same questions:
AI RISK TOOL VENDOR QUESTIONNAIRE - FDA pathway and authorization number; cleared indications - Supported modalities: 2D FFDM, DBT synthetic 2D, vendor/model of mammography units - Deployment: on-prem, cloud, hybrid; PHI handling; BAA provided? SOC 2 report? - Integration: DICOM SR / structured output? HL7 ORU to RIS? PowerScribe/reporting fields? - Turnaround time per exam; behavior when exam fails QC - Performance by race/ethnicity, age band, and breast density category - Pricing model: per-study, annual license, revenue share - Existing reimbursement support or billing guidance - Customer references at practices of similar volume - Contract terms: data ownership, model updates, exit/data return -
Define the risk pathway before go-live. Write down exactly what happens at each score band and get sign-off from radiology leadership and referring physicians. A starting template:
RISK PATHWAY (draft - requires medical director approval) Average risk band: - Routine annual/biennial screening per current guideline - Score included in report; standard result letter Elevated risk band: - Report language recommending risk discussion with PCP - Offer high-risk clinic consult; consider supplemental imaging if dense High risk band: - Recommend supplemental breast MRI (or contrast-enhanced mammography if MRI contraindicated) - Refer for genetic counseling evaluation - Navigator outreach within 10 business days; track completion Escalation owner: [Name, role] Audit cadence: quarterly -
Map the data flow with IT. The practical question is where the score lands. If it never reaches the radiologist’s report and the patient letter, it’s useless. Here is a typical routing rule, shown as a generic illustration rather than any vendor’s actual config:
# Illustrative DICOM routing rule (adapt to your PACS/router) rule: send_screening_mammo_to_risk_ai match: Modality: MG StudyDescription: ["*SCREEN*", "*SCRN*"] ImageType: ["ORIGINAL\\PRIMARY"] exclude: StudyDescription: ["*DIAG*"] destination: RISK_AI_NODE # AE title provided by vendor return: format: DICOM_SR forward_to: [PACS_MAIN, REPORTING_SYSTEM] hl7_oru_to: RIS alert_on_failure: radiology-it@yourpractice.com -
Prepare patient communication. A “high risk” letter will surprise patients with no family history. Draft the language carefully and have clinical and legal teams review it:
Write a patient result letter (6th-grade reading level, under 200 words) for a woman whose screening mammogram was normal but whose AI-based 5-year risk assessment placed her in a higher-risk category. Requirements: - Lead with: no cancer was found on today's mammogram - Explain that the risk estimate looks at patterns in breast tissue - Do NOT state a diagnosis or imply cancer is present - Recommend talking with her doctor about extra screening options - Include a phone number placeholder for our breast health navigator - Calm, direct tone; no jargon -
Build the billing and revenue model. Model three scenarios: absorbed cost as a competitive differentiator, a self-pay add-on, and a bundled high-risk program. Count the downstream MRI, ultrasound, and genetic counseling volume the score generates, because most of the revenue sits there. Talk to payers early. Some value-based contracts will pay for prevention that lowers late-stage treatment costs.
-
Audit outcomes from day one. Track the share of patients flagged high risk, MRI completion rate, supplemental cancer detection rate, and patient complaints. Those numbers strengthen payer negotiations and the marketing story.
How it compares
The “Clairity vs Lunit vs iCAD” question comes up in every evaluation. These tools mostly do different jobs, so many practices will run more than one.
| Tool | Primary job | Risk horizon | U.S. regulatory status (risk use) | Best fit |
|---|---|---|---|---|
| Clairity Breast | Image-only future cancer risk prediction | 5 years | FDA De Novo authorized (first in category) | Practices building high-risk and supplemental screening programs |
| iCAD ProFound AI (Detection + Risk) | Cancer detection on DBT/2D; separate short-term risk module | Detection now; risk module estimates near-term (about 2-year) risk | Detection FDA cleared; check current status of the risk module before relying on it | Practices wanting detection plus risk from one vendor (iCAD is now part of RadNet) |
| Lunit INSIGHT MMG / DBT (incl. Volpara) | Cancer detection; density and quality via Volpara | Current exam; density feeds risk models | Detection FDA cleared; density tools cleared | High-volume reading groups focused on detection accuracy and workflow |
| Questionnaire models (Tyrer-Cuzick, Gail) | Risk from history and demographics | 5-10 years and lifetime | Clinical standard, not a device | Guideline-based MRI eligibility and genetics referral today |
Don’t treat image-based and questionnaire-based risk as either/or yet. Guidelines and payers still reference lifetime risk from traditional models for MRI eligibility. The smart move is to run both, use Clairity to catch women the questionnaire misses, and document where the two disagree.
What’s next
Reimbursement is the biggest variable. Watch for a Category III CPT code application, CMS coverage decisions, and the first commercial payers to cover AI risk scoring, most likely inside value-based or high-risk program contracts. Once a code arrives, adoption will accelerate. Until then, the economics depend on downstream services and competitive positioning.
Second, watch guidelines. If the American College of Radiology or other societies start citing image-based risk in their supplemental screening recommendations, the risk score becomes standard of care rather than an extra. Prospective outcome data matters here too. Retrospective validation got Clairity authorized, but guideline committees want prospective evidence that more high-risk women get MRI and that cancers are caught earlier.
Third, expect competition and bundling. Clairity’s De Novo created a predicate, so other vendors now have a clearer 510(k) path for their own risk products. Detection vendors will add risk modules, risk vendors will partner with PACS and reporting platforms, and large radiology groups will negotiate enterprise AI bundles. For business owners, that means pricing leverage over the next 12-24 months. Don’t sign long exclusive contracts too early.
Frequently Asked Questions
What is Clairity Breast?
Clairity Breast is an FDA-authorized AI platform that analyzes a routine screening mammogram and predicts a woman’s risk of developing breast cancer over the next five years. It was the first device authorized for this purpose and uses image patterns rather than questionnaire answers.
How is 5-year breast cancer risk prediction different from AI cancer detection?
Detection AI flags suspicious findings on the current exam to help radiologists catch existing cancers. 5-year breast cancer risk prediction estimates the chance cancer develops in the future, even when today’s mammogram is completely normal. The outputs lead to different actions: diagnostic workup for detection, and prevention and supplemental screening for risk.
Does Clairity replace Tyrer-Cuzick or other risk questionnaires?
Not yet. Many guidelines and payer policies still use lifetime risk from traditional models to decide MRI eligibility. Most practices will run image-based and questionnaire-based risk side by side and use the AI score to find high-risk women the questionnaire misses.
Can imaging centers bill for an AI mammogram breast cancer risk score?
No widely reimbursed, dedicated code exists for this service yet. Practices use self-pay add-ons, bundled screening packages, and value-based contracts, and they count on the downstream MRI, ultrasound, and genetic counseling volume. Check with your billing team and payers before launch.
Is FDA authorized breast cancer AI safe to use on women with dense breasts?
Dense tissue is one of the main reasons image-based risk is useful. Density both raises risk and hides cancers on mammography. Check the FDA decision summary for the populations and exam types the device was validated on, and set up a clear dense breast supplemental screening pathway for patients who score high.
What should a practice do first if it wants to adopt Clairity?
Start with clinical pathway design and the revenue model, not the software contract. Decide what happens at each risk level, who owns follow-up, how patients are told, and how the program pays for itself. Then evaluate vendors against that plan.
Go deeper than this article
This article covers the essentials. Our premium eguide library gives you the full step-by-step playbooks — prompts, workflows, and copy-paste recipes you can put to work today.
