MONA
An AI software as a medical device that reads fundus photographs to flag diabetic retinopathy and diabetic maculopathy. A nurse or technician can run the screening, and a referral decision comes back within seconds instead of weeks.
What is MONA?
MONA is an artificial intelligence software as a medical device built by MONA.health bv, a Belgian company based in Bertem, near Leuven, and backed by KU Leuven and the Health unit of VITO, the Flemish research and technology agency. The product deliberately defines itself by what it is not: it is neither a retinal camera nor software embedded inside one. It is an analysis layer that accepts pictures from most fundus cameras already in service, which means a practice can adopt it without replacing its imaging equipment.
The clinical problem it targets is access rather than accuracy. Screening for diabetic retinopathy has traditionally required an ophthalmologist to read the images, and with a limited number of specialists in any given area, waiting lists grow while a preventable disease progresses. MONA moves the reading step to an algorithm, so the capture can be performed by a nurse or a technician. The vendor puts the whole patient encounter at no more than three minutes, with the referral decision returned in seconds.
On performance, the company publishes figures rather than adjectives. For referable diabetic retinopathy it claims a sensitivity of 90% with a specificity of 95%, rising to 99% sensitivity for non-proliferative and proliferative disease. For hard exudates predictive of macular oedema it reports 91% sensitivity and 94% specificity. The deep learning models behind these numbers were trained, tested and validated against more than 273,000 retinal images.
A second, separately documented product, MONA GLC, predicts whether someone should be referred into primary open-angle glaucoma screening, with sensitivity and specificity both above 85% and overall accuracy above 96% in an evaluation population of 11,952. Its instructions for use, complete with a unique device identifier, are published on the site.
The solution combines an optional hardware bundle, a laptop with a camera, and a web application; the web component is the regulated medical device. Use is restricted to healthcare professionals, and the software is explicitly not intended to support diagnostic or therapeutic decisions. It routes patients into established care pathways, and nothing more.
What it does
- Screens colour fundus photographs for referable diabetic retinopathy
- Detects diabetic maculopathy through hard exudates predictive of macular oedema
- Returns a referral decision within seconds of the images being uploaded
- Predicts the need for referral into open-angle glaucoma screening via the MONA GLC product
- Generates a full report inside the web application and shares it with the ophthalmologist
- Lets the healthcare professional add diagnosis-related annotations to a patient record
- Works with images from most existing fundus cameras rather than requiring new hardware
When to use MONA / When not to
A quick filter to help you decide if MONA is the right fit.
When to use MONA
- Ophthalmology practices whose screening slots are the bottleneck holding up everything else
- Diabetologists and endocrinologists who want to add an eye check to a routine diabetes consultation
- Primary care and general practices running the annual diabetic eye review
- Screening services where nurses or imaging technicians, not doctors, capture the pictures
- Hospitals and health systems trying to raise screening uptake without hiring more specialists
When not to use MONA
- Patients themselves: the terms restrict use to registered healthcare professionals
- Anyone expecting a diagnosis, since the software only signals whether a referral is warranted
- Screening people under 18, wearers of prosthetic eyes, or anyone contraindicated for fundus imaging
- Monitoring the progression of an already diagnosed glaucoma, which is outside the stated intended use
- Buyers who need a published price before opening a conversation, as everything is contract-based
How to use MONA
A typical end-to-end flow, from setup to results.
- Agree the commercial terms, since access runs through a contract with the practice, clinic or hospital
- Accept the terms of use and the privacy policy, which is a precondition for using the service
- Use an existing fundus camera, or the optional laptop-and-camera bundle supplied by MONA.health
- Have a nurse or technician capture one colour retinal image of each eye during the annual review
- Follow the image requirements: in-focus, unaltered camera exports, centred on the optic disc, 45-degree field
- Upload the images to the MONA web platform over an internet connection
- Attach the images to an existing patient record or create a new one
- Let the deep learning models analyse the images and return a referability estimate within seconds
- Read the outcome: refer the patient to an eye specialist, or confirm the next annual exam
- Add any diagnosis-related annotations, then share the generated report with the ophthalmologist
Pros & Cons
Pros
- Shifts screening from the ophthalmologist to a nurse or technician, which is where the bottleneck actually sits
- Referral decision in seconds and a three-minute patient encounter
- Works with the cameras a practice already owns, so there is no forced hardware purchase
- Publishes concrete sensitivity and specificity figures instead of vague accuracy claims
- Model validation rests on an unusually large image set, more than 273,000 retinal photographs
- Published instructions for use and a unique device identifier for the MONA GLC product
- Personal data is stated to sit on Amazon servers inside the European Union, under Belgian law
Cons
- No published price anywhere: the fee is set in a negotiated agreement, so no benchmark is possible
- Neither a free trial nor a free plan is mentioned, and there is no self-service entry point
- Contractual liability is capped at 2,000 EUR, which is modest for a regulated medical device
- No public API and no technical documentation for integrating results into an existing record system
- Published instructions for use cover only MONA GLC; the diabetic retinopathy product has none online
- No information security certification such as ISO 27001 or SOC 2 is claimed or evidenced
- The marketing site is visibly under-maintained, with an empty disclaimer page and a 2021 copyright
Pricing & Plans
MONA.health does not publish any price. The terms of use state only that the services are made available to users against payment of a fee, with the amount set in the main agreement concluded with the practitioner, the medical practice or the hospital. There is no pricing page, no published tier and no rate card; a full enumeration of the site's sitemap confirms the absence rather than merely failing to find one. No free plan and no free trial are mentioned. Readers should note that the only euro amount on the site, 2,000 EUR, is the contractual liability cap and not a price. Obtaining a figure requires contacting the company.
Data, GDPR & hosting
A consolidated view of how MONA handles your data.
GDPR overview
Implementation is concrete rather than declarative. The privacy policy names Regulation (EU) 2016/679 explicitly, identifies MONA.health bv as data controller with its registered office and enterprise number, and sets out a legal basis for each processing purpose. All the individual rights are listed: access and copy, rectification, erasure, restriction, objection, portability, withdrawal of consent, and the right not to be subject to a purely automated decision. Rights are exercised through a dedicated address, privacy@mona.health, or by post to the registered office. Retention periods are published per data category, breach notification is promised without undue delay, and Belgian law governs with the Antwerp courts named as competent. No data protection officer is designated, and no Article 27 representative appears, which is consistent with an editor established inside the European Union.
Who owns the data?
The terms are unusually clear on this point: you keep ownership of your data at all times, while MONA.health keeps ownership of the software and every intellectual property right attached to it. The catch sits in clause 4.4. By uploading content you grant MONA.health a non-exclusive, royalty-free, worldwide licence to use, copy, store, modify, transmit and display it in order to supplement and update the service, and that licence is both sub-licensable and transferable. Ownership therefore stays with you, but the rights you hand over are broad and are not limited to your own account. For the website itself, MONA.health acts as data controller.
Reuse rights
Nothing in the terms stops a healthcare professional from reusing the images and results they generate: the customer remains the owner and the licence granted to MONA.health is non-exclusive, so no permission has to be requested to keep using one's own clinical material. What the terms do restrict is the software, not the data. Reverse engineering, resale, sublicensing, renting out access, integrating the solution into other software without written consent and bulk automated collection are all prohibited. On MONA.health's side, personal data is stated to serve only the delivery, communication and improvement of the service, with transfers to third parties limited to what a contract strictly requires and covered by a signed processing agreement. Whether that improvement extends to training AI models is never spelled out.
Data retention & training
Hosting summary
The privacy policy states that personal data is stored on high-security Amazon servers within the European Union. No individual country is named, so the site commits to a region rather than a jurisdiction. Two independent technical observations made during collection point the same way: the application host resolves to an Amazon load balancer in the eu-central-1 region in Frankfurt, and one page of the site leaks a canonical URL pointing at a raw EC2 hostname in that same region. The public marketing site itself sits behind Cloudflare, whose address is anycast and therefore says nothing about where anything is actually stored. Governing law is Belgian and the Antwerp courts are named as competent, with the Belgian Data Protection Authority as supervisory authority. Stated safeguards include physical and digital access controls, password management, internal security policies and secure storage. No list of subprocessors is published, and no data processing agreement is offered to customers on the site.
Things to keep in mind
Risks and trade-offs to weigh before adopting MONA.
- Automation bias is the central risk: the vendor states that false negatives occur, so a reassuring result must never override a patient reporting vision problems
- The uploaded-content licence in clause 4.4 is broad, sub-licensable and transferable, and the site never states whether images are used to train the models
- Contractual liability is capped at 2,000 EUR, which is low for a device involved in referral decisions about sight loss
- Published instructions for use exist only for MONA GLC; the diabetic retinopathy product promoted on the homepage has no equivalent document online
- The CE marking appears only in a symbol legend, with no notified body named and no certificate number given
- Deskilling is a real long-term risk if practices stop maintaining in-house reading competence for retinal images
- The site is poorly maintained, with an empty disclaimer page, a job page with no openings and a 2021 copyright, which complicates due diligence
Setup & Integrations
Technical difficulty
Low for the end user. Screening runs in a web application, so nothing is installed locally beyond a browser and an internet connection, and the practice can keep its existing fundus camera or take the optional laptop-and-camera bundle. The real effort is contractual rather than technical: access comes through a negotiated agreement, and the terms and privacy policy must be accepted first. MONA GLC installs on a workstation whose requirements sit in an engineering manual that is not published, and the vendor recommends a strong password, active antivirus and firewall, and current security patches.
Deployment
Behind MONA
Social
Resources
All the official URLs gathered for verification and reference.
Frequently asked questions
Does MONA replace the ophthalmologist?
Who is allowed to use it?
Do we have to buy a specific camera?
How long does a screening take?
Which conditions does it cover?
How accurate is it?
What does it cost?
Where is the data hosted?
Is there an API?
Is there a mobile app?
Should you pick MONA?
MONA is a narrow, seriously regulated product rather than a general-purpose AI assistant, and it should be judged on that basis. Its value proposition is organisational before it is technical: by letting a nurse or a technician capture the images and having an algorithm decide who needs referral, it attacks the specialist bottleneck that makes diabetic eye screening slow and, too often, skipped. The published performance figures are specific, the validation set is large, and the company is transparent enough to state plainly what the software must not be used for.
The reservations are mostly about disclosure rather than substance. Pricing is entirely opaque: everything runs through a negotiated agreement, so a prospective buyer cannot form even a rough idea of cost without contacting the company. Documentation is uneven, with full instructions for use and a device identifier published for the glaucoma product but nothing equivalent for the diabetic retinopathy offering that the homepage leads with. The 2,000 EUR liability cap is low for a medical device, and no information security certification is evidenced. The marketing site itself is visibly neglected, which says little about the software but does make due diligence harder than it should be.
For an ophthalmology practice, a diabetes clinic or a screening service that already owns fundus cameras and is constrained by reading capacity, MONA is worth a conversation. Expect that conversation to be a procurement process rather than a sign-up, and expect to ask directly for the regulatory file, the security posture and a written position on whether uploaded images may be used to improve the models.
- Choosing a selection results in a full page refresh.
- Opens in a new window.