Ophthalmology · Handheld fundus camera for AI diabetic-retinopathy screening
Optomed Aurora
Optomed (camera); AEYE Health (AEYE-DS AI)
The Optomed Aurora sits at an unusually clear illustration of how AI medical devices are actually assembled: a hardware camera from one company, an autonomous diagnostic algorithm from another, cleared and marketed together. The camera carries its own FDA clearance as ophthalmic imaging hardware; the headline capability — autonomous, on-the-spot detection of more-than-mild diabetic retinopathy from a single image per eye — comes from AEYE-DS, a separately cleared device legally held by AEYE Health, not Optomed. That attribution is the single most important thing a procurement reviewer should carry from this entry. The evidence for AI screening on Aurora images is genuinely encouraging and, notably, partly independent: academic groups in Italy, Turkey and at Harvard's Joslin Diabetes Center have reported strong sensitivity and specificity across several different algorithms. But the regulatory footprint for the autonomous AI is US-only via AEYE Health, the camera's own EU and Canadian approvals could not be pinned to specific register numbers on this review, and the evidence is diagnostic-accuracy validation rather than outcome trials. A single mark reflects a real, well-evidenced screening platform whose strongest claims belong, in part, to a different manufacturer.
Performance Metrics
Clinical Evidence
The evidence for AI diabetic-retinopathy screening on Optomed Aurora images is a mix of independent academic validation and manufacturer-affiliated work, and it spans several different algorithms — which is both a strength (the camera is not tied to one vendor's numbers) and a reason to read each result against its specific algorithm and authorship. The strongest independent evidence comes from two academic studies. Lupidi and colleagues (Acta Diabetologica 2023; PMID 37154944) validated the built-in SELENA+ algorithm on the Aurora in a real-life screening setting of 256 consecutive patients (256 eyes), reporting sensitivity and specificity of 96.8% each for any diabetic retinopathy and a weighted kappa of 0.935 against a retina specialist's dilated examination. An independent Turkish head-to-head study (Eye (London) 2024; PMID 38467864) evaluated three non-mydriatic cameras with the EyeCheckup AI across 900 patients at an endocrinology clinic and reported Aurora sensitivity of 90.5% and specificity of 97.2% for more-than-mild retinopathy. A further independent study from the Joslin Diabetes Center / Beetham Eye Institute at Harvard (British Journal of Ophthalmology, 2024; PMID 37094836) compared one-, two- and five-field handheld imaging against standard seven-field ETDRS photography, finding — as expected — that more imaging fields improved referable-DR performance; the exact five-field Aurora sensitivity and specificity figures were not re-verified against the full text on this review and are therefore not stated as precise values here. The manufacturer-affiliated evidence is consistent in direction. A real-world evaluation of the Aireen algorithm on the Aurora (Diabetes Technology & Therapeutics 2025; PMID 40824290; Optomed-affiliated authors) reported sensitivity 94.8% and specificity 91.4% across 624 fundus images, and a 21-algorithm comparison on Aurora images (Annals of Medicine 2024; PMID 38738798; 156 patients) documented wide between-algorithm variation (mean agreement with ophthalmologist grading 79.4%, range roughly 49–92%) — a useful reminder that camera and algorithm must be evaluated as a pair. For the US autonomous pathway specifically, the AEYE-DS registrational evidence lives in the FDA 510(k) summaries rather than the indexed literature: two prospective, multi-centre, single-arm, blinded studies (roughly 317 and 362 participants) using the Aurora reported sensitivity in the 92–93% range and specificity in the 89–94% range against ETDRS reading-centre grading, with imageability near 99%. No randomised outcome trial of screening-on-Aurora was located; the evidence establishes diagnostic accuracy, not a change in patient outcomes.
| Study | Design | n | Sensitivity | Specificity | AUC | Published |
|---|---|---|---|---|---|---|
| Lupidi M, et al. (independent; University of Perugia / Italy) | ProspectiveProspective | 256 | 96.8% | 96.8% | Weighted kappa 0.935 vs retina-specialist dilated exam (any DR) | Acta Diabetol, 2023; PMID 37154944 — SELENA+ on Optomed Aurora, 256 consecutive patients |
| Head-to-head 3-camera study (independent; Akdeniz University, Turkey) | ProspectiveProspective | 900 | 90.5% | 97.2% | mtmDR; EyeCheckup AI across Canon / Topcon / Optomed Aurora | Eye (Lond), 2024; PMID 38467864 |
| AEYE-DS registrational program (AEYE Health; prospective, multi-centre, single-arm, blinded) | ProspectiveProspective | 362 | 92–93% | 89–94% | vs ETDRS reading-centre grading; imageability ~99% (per FDA 510(k) summaries, K221183 / K240058) | FDA 510(k) summaries; NCT05857943 (registrational data not separately indexed as a paper) |
| Aireen real-world evaluation (Optomed-affiliated) | RetrospectiveRetrospective | 624 | 94.8% | 91.4% | Accuracy 92.7%; Aireen on Optomed Aurora | Diabetes Technol Ther, 2025; PMID 40824290 (manufacturer-affiliated) |
Clinical Pulse
Verified clinician sentiment
Verified clinician reviews launching soon · Apply to contribute →
Reviews are signed by clinicians verified against their regulator's public register.
Inside the algorithm
Editorial featureHow a handheld camera and a third-party AI screen for retinopathy.
Five stages — from raw input to verdict — drawn from manufacturer documentation and the public regulatory record.
- INGEST
- NORMALISE
- DETECT
- LOCALISE
- VERDICT
Stage 01 · INGEST
A non-specialist captures a fundus image at the point of care.
The Optomed Aurora is a portable, non-mydriatic retinal camera small enough to be used outside an eye clinic — in primary care, endocrinology, a pharmacy or a mobile setting. A trained but non-specialist operator captures a fundus image, typically without pharmacological dilation.
The camera is the image-capture front end only; it contains no autonomous diagnostic algorithm. What happens to the image next depends on the software paired with it in that market.
Input
Colour fundus image
Inference
Performed by paired AI, not the camera
Inside the Auris+ Listing
Five more sections complete this device’s Auris+ Listing.
Decision Ledger
ProPro unlocks a private, cross-vendor log of every case you read on this device — what the AI called, what you concluded, and a one-line reason if you overrode it. Ready for the EU AI Act's deployer logging obligations when they land in 2028.
Clinical Evidence Deep Dive
ProPro unlocks the structured clinical-evidence summary — study count, target patient population, and a tabular accuracy-metrics view drawn from peer-reviewed sources.
Peer-Reviewed Publications
ProPro unlocks the curated peer-reviewed publication list with PubMed cross-links — the citation backbone of every editorial verdict.
Post-Market & Regulatory Conditions
ProPro unlocks the post-market surveillance summary, recall record, and the conditions of approval that bound real-world use.
AI Algorithm Version History
ProPro unlocks the chronological record of algorithm version changes — what changed when, drawn from manufacturer changelogs and regulatory filings.
Regulatory Approvals
K240058
Class II
104201
Safety Record
No Optomed recalls or FDA safety communications for the Aurora camera were identified in publicly available reporting as of July 2026. For the AEYE-DS AI, the source dossier recorded a single MAUDE entry (report number MW5165159, dated 2025-01-22, categorised "Injury", alleging the software failed to return a complete diagnosis). This routine could not corroborate that specific report against the primary MAUDE record — the FDA MAUDE and openFDA databases return 403 from this environment and the report number did not surface in any web index — so it is carried here only as an unconfirmed, uninvestigated lead, not as an established failure or recall, and should be confirmed directly against accessdata.fda.gov before any weight is placed on it. The residual risk profile of the pairing is that of any autonomous screening tool: a false-negative result could defer a needed referral, which is why the AI is indicated for screening (not diagnosis) and positive or ungradable results route to full ophthalmological evaluation.
Intended Use & Indications
The Optomed Aurora is a portable, non-mydriatic retinal camera cleared by the FDA as ophthalmic imaging hardware. It is designed for use outside traditional eye-care settings — primary care, endocrinology, pharmacies and low-resource clinics — where its handheld form factor lets a non-specialist capture fundus images. The camera on its own performs no diagnosis. Its diabetic-retinopathy screening capability is supplied by separately regulated AI software: in the United States, the "Aurora AEYE" configuration pairs the camera with AEYE-DS, an autonomous AI diagnostic system that analyses one image per eye and returns a binary more-than-mild-diabetic-retinopathy result without a clinician interpreting the image. AEYE-DS is a distinct FDA-cleared device legally held by AEYE Health, Inc.; the Aurora camera clearance and the AEYE-DS AI clearance are separate regulatory products, and the AI authorisation belongs to AEYE Health. Independent research groups have additionally validated other algorithms — EyeRIS SELENA+, EyeCheckup and Aireen among them — on images captured with the Aurora. This entry treats the Aurora as the imaging platform for AI screening while attributing each AI clearance to its actual holder.