AI Diabetic Retinopathy Screening for Optometrists: What the 2025 FDA Clearance and CPT 92229 Mean for Your Practice
AI Diabetic Retinopathy Screening for Optometrists: What the FDA Clearance and CPT 92229 Mean for Your Practice
Autonomous AI diabetic retinopathy screening has crossed three thresholds at once: regulatory clearance, clinical-grade accuracy, and reimbursable billing. For optometry practices that co-manage diabetic patients or sit near primary care hubs, 2025 is the year this stops being a future technology and starts being a clinical and financial decision.
This post breaks down what has actually changed, what the evidence says, and what adding autonomous AI DR screening to your practice looks like in practice — including the billing mechanics most optometrists haven't seen explained clearly.
What Changed: Three Enablers Finally Converged
AI-based diabetic retinopathy screening is not new. What is new in 2025 is the simultaneous arrival of three things that individually meant nothing but together make it viable:
1. FDA De Novo clearance for fully autonomous AI reads AEYE Health received FDA clearance for the first fully autonomous AI diabetic retinopathy screening system that requires only a single handheld image per eye — no specialist review required for a result. The AI reads the image and returns a clinical decision autonomously.
2. Portable, non-mydriatic hardware that works without dilation Earlier AI DR systems required tabletop fundus cameras and a dilated pupil. Current FDA-cleared systems work with handheld, non-mydriatic cameras, achieving >99% imageability in undilated patients. A camera that can go into any room eliminates the workflow barrier that kept DR screening siloed in specialist clinics.
3. CPT code 92229 — a dedicated billing code for autonomous AI DR screening CPT 92229 (Imaging of retina for detection or monitoring of disease; with point-of-care automated analysis and report, unilateral or bilateral) specifically covers autonomous AI retinal imaging for DR screening. Medicare covers it. Without this code, the clinical case existed but the financial case didn't. Now both do.
What the Clinical Evidence Shows
AEYE Health's FDA-cleared system was validated across a diverse population and published in peer-reviewed literature:
| Metric | Result |
|---|---|
| Sensitivity (detecting referable DR) | 92–93% |
| Specificity | 89–94% |
| Imageability (undilated) | >99% |
| Images required per eye | 1 |
| Time from image to result | Minutes |
| Specialist review required | No |
For context: the American Diabetes Association's minimum sensitivity threshold for autonomous AI DR screening is 85%. The AEYE system exceeds that threshold meaningfully, which is part of why the FDA granted De Novo clearance rather than requiring a full PMA pathway.
The >99% imageability figure is clinically significant. Previous AI DR systems failed to produce usable images in 5–15% of attempts due to media opacity, pupil size, or patient cooperation. One failed image means a return visit, which in a diabetic population often means no follow-up at all.
What "Autonomous" Actually Means
Autonomous does not mean unsupervised in the medical liability sense. It means:
- The AI produces a binary clinical result (referable / not referable) without requiring a specialist to view and sign off on each image
- That result is what is documented and acted upon
- The optometrist is still the ordering and supervising clinician
You are not replaced. You are the practitioner who orders the test, receives the report, counsels the patient, and makes the referral decision. The AI handles the image interpretation — the same way a lab machine interprets a blood glucose reading and returns a number, which you then act on.
The Billing Mechanics: CPT 92229 in Detail
What the code covers
CPT 92229 covers:
- Point-of-care retinal imaging
- Automated (AI) analysis of that image
- A structured report
- Bilateral or unilateral coverage
The code applies when the imaging is performed and the AI analysis is produced at the same encounter. It is distinct from CPT 92228 (remote imaging for DR detection with physician interpretation) — 92229 is specifically for autonomous AI reads at the point of care.
Who can bill it
Optometrists can bill CPT 92229 where their scope of practice permits retinal imaging for disease detection. Scope varies by jurisdiction — confirm with your local optometry board before billing, but most states and countries where optometrists perform fundus photography already fall within scope.
What payers cover it
Medicare covers CPT 92229. Commercial payer coverage is expanding but still inconsistent. Run a pre-authorisation check with your top five payers before investing in equipment. Several commercial plans now cover it, particularly those with value-based care arrangements around diabetes management.
What you cannot double-bill
CPT 92229 cannot be billed on the same day as CPT 92228 or in combination with other retinal imaging codes for the same encounter. It is also not billable in addition to a comprehensive dilated fundus exam (92004/92014) if a full exam was also performed — document clearly which service was rendered.
The Practice Workflow
Here is what an AI DR screening workflow actually looks like integrated into an optometry practice:
Identifying eligible patients
Your practice management system should flag diabetic patients automatically at check-in. Any patient with a diabetes diagnosis in their record who has not had a DR screen in the past 12 months is a candidate. This is where your EHR or PMS is doing the work — optometrists who track diagnoses systematically will capture far more eligible patients than those running manual recall.
The encounter
- Patient arrives (can be a standalone screening visit or added to a routine exam)
- Technician or front desk staff captures one non-mydriatic fundus image per eye using the handheld camera — takes 3–5 minutes
- Image uploads automatically to the AI analysis platform
- AI returns a report: referable DR (yes/no) plus severity grading if referable
- Optometrist reviews the report, counsels the patient, and documents the result
- If referable: refer to ophthalmology. If not referable: schedule next annual screen
Documentation requirements
For CPT 92229 claims to hold up on audit:
- Document the diabetic diagnosis (ICD-10 E11.x or E10.x as appropriate)
- Document that autonomous AI analysis was used and name the system
- Include the AI-generated report in the patient record
- Document your clinical interpretation and next steps
The Business Case
Revenue per encounter
CPT 92229 Medicare reimbursement rates vary by geography (use the CMS Physician Fee Schedule for your MAC's rates). As a rough benchmark, practices are seeing $30–60 per encounter in Medicare reimbursement depending on region. At 2–3 diabetic patients per clinic day, this adds up.
But the bigger financial case is downstream:
| Scenario | Impact |
|---|---|
| Detecting referable DR earlier | Better patient outcome, reduced liability exposure |
| Generating a referral to ophthalmology | Strengthens co-management relationships |
| Annual recall billing | Each screened patient = annual recall visit |
| Adding a screening visit type | Revenue from patients who would otherwise only come for glasses |
| Differentiating from competitors | Attract diabetic patients from primary care referrals |
Equipment cost
Handheld non-mydriatic fundus cameras capable of producing AI DR-compatible images range from approximately $8,000–$20,000. Some AI platform providers offer equipment through a per-use or subscription model that eliminates upfront capital cost. Run your break-even analysis against your diabetic patient volume before purchasing outright.
What This Means for the Broader Direction of Optometry
AI DR screening is the first AI diagnostic workflow in optometry to clear all three bars — clinical evidence, regulatory approval, and reimbursement — simultaneously. It will not be the last.
The underlying pattern is important: AI handles the pattern-recognition step (grading an image), while the optometrist handles everything else (ordering, counselling, referring, managing the patient relationship). This is not a threat to optometry scope. It is an expansion of what a single optometrist can accomplish in a day without requiring specialist backup for every image read.
Practices that integrate this workflow now will have established referral relationships, trained staff, and billing infrastructure by the time AI-assisted glaucoma screening, AMD detection, and neurological screening tools follow the same path through the FDA.
Why CarrotByte
Realising the value of AI DR screening depends on identifying eligible patients consistently, not just occasionally. CarrotByte's patient management system lets you tag diabetic diagnoses, set automated recall triggers for annual DR screens, and track which patients have been screened versus which are overdue — so no eligible patient slips through because someone forgot to check.
Managing a new billable service also means documenting it correctly every time. CarrotByte's encounter records support structured documentation that holds up under payer audit, and the recall system ensures your AI DR screening programme runs year-round without manual list management.
See how CarrotByte supports diabetic patient management →