An imaging AI marketplace is the layer that sits between a radiology department and the fast-growing catalog of algorithms that read scans: a single storefront, integration, and viewer through which a hospital can switch on one vendor's stroke-triage tool and another's lung-nodule detector without hand-wiring each to the PACS. This page sets the main platforms beside one another on the attributes that genuinely differ — where the software runs, whose algorithms it carries, and which regulator cleared them — and ties every cell to the platform's own documentation or the FDA's record. It ranks nothing, and it names no winner. As of July 2026.
The regulatory floor every marketplace sits on
Before comparing storefronts, fix the thing they all share. The FDA's AI-Enabled Medical Device List is the closest thing the field has to a census of authorized imaging AI 1, and the specialty split is lopsided: radiology accounted for 74.4% of the 168 machine-learning-enabled devices the FDA authorized in 2024 8. The American College of Radiology's Data Science Institute keeps the practitioner-facing index — its AI Central directory has curated more than 200 FDA-cleared imaging AI products from more than 100 manufacturers 2.
The load-bearing fact for everything below: a clearance attaches to an algorithm, not to a marketplace. When you read that a platform "offers 80 apps," the regulatory authorization sits on each of those apps individually, granted to its developer as Software as a Medical Device. The orchestration layer — the viewer, the routing, the worklist integration — is generally not the cleared diagnostic device. A handful of vendors both build their own cleared algorithms and run the platform; most curate third-party apps. That distinction is the first column of the table.
One more piece of context sets expectations. A peer-reviewed taxonomy of 1,016 FDA authorizations found that quantitative image analysis is the most common device function and that none of the authorized devices yet use large language models 7. Whatever a marketplace says about "foundation models," the cleared products on it measure pixels; they do not chat.
The platforms, attribute by attribute
The table below is a snapshot of documented attributes, each cell traceable to the platform's own materials or the FDA record. Counts of available applications are vendor-stated and move month to month; treat them as order-of-magnitude, not as a scoreboard. As of July 2026. Clearances, features, and availability change — reconfirm against the vendor and the FDA list before you deploy.
| Platform | Model origin | Where it runs | Clearance basis of the apps | Documented breadth (vendor-stated) | Source |
|---|---|---|---|---|---|
| Aidoc aiOS | First-party algorithms plus platform | Deployed into the health system; foundation-model-scale management | FDA-cleared indications held by Aidoc; one 2026 clearance combines 14 body-CT indications | "More than 100 million patient cases analyzed" | 3 |
| Nuance / Microsoft Precision Imaging Network | Third-party aggregation | Built on Microsoft Azure | Each third-party model carries its own clearance | "35+ third-party medical imaging AI models from over 20 vendors" | 4 |
| deepc deepcOS | Third-party aggregation (vendor-inclusive) | Cloud-native | External regulatory-approved partners; devices cleared for the EU under MDR/MDD | Aggregates external partner solutions via one integration | 5 |
| Sectra Amplifier Marketplace | Third-party aggregation, tied to Sectra viewer | Integrated with the Sectra diagnostic viewer | Every listed app carries CE marking, FDA clearance and/or Health Canada approval | Catalog integrated at the point of care | 6 |
| ACR AI Central (reference directory rather than a marketplace) | Neutral catalog | Web directory | FDA-cleared only | "More than 200" products from "more than 100" manufacturers | 2 |
A few notes make the columns readable rather than reductive.
Model origin is the sharpest operational divide. Aidoc develops its own cleared algorithms and runs aiOS as the deployment and governance layer around them; the company states its platform has analyzed more than 100 million patient cases and that a 2026 clearance folds 14 indications (11 newly cleared plus 3 prior) into a single body-CT triage solution, reporting a mean sensitivity of 97% and specificity of 98% across the new indications 3. By contrast, Precision Imaging Network, deepcOS, and the Sectra Amplifier Marketplace are principally aggregators: they curate and route algorithms built by other companies, each of which arrives with its own clearance. Neither pattern is superior in the abstract — a first-party stack can mean tighter integration and a single accountable vendor, while an aggregator can mean wider algorithm choice and less lock-in. They are different trades, not different tiers.
Where it runs and how it clears interact with your jurisdiction. Precision Imaging Network is documented as running on Azure with 35+ models from over 20 vendors 4; deepcOS states that its results come from external regulatory-approved partners and that all devices are cleared for the EU under MDR/MDD — an EU-first posture, with a narrower FDA-cleared subset for US use 5. Sectra's Amplifier Marketplace requires that every listed application be validated for point-of-care use through CE marking, FDA clearance and/or Health Canada approval and integrate with the Sectra viewer 6. If you practise under one regulator, the relevant question is not "how many apps" but "how many of the apps I need hold the clearance my jurisdiction recognizes."
What the evidence does — and does not — settle
A marketplace comparison can tell you what each platform carries and how it plugs in. It cannot tell you how a given algorithm will perform on your patients, and the published record is candid about why that gap persists.
Transparency lags approval. Among the 2024 FDA machine-learning authorizations, only 29.2% reported both sensitivity and specificity, just 15.5% included demographic data on the study population, and a predetermined change control plan appeared in only 16.7% of summaries 8. A clearance confirms the FDA's review threshold was met; it does not hand you a full performance-and-equity profile, and a marketplace listing rarely adds one. This is why external validation in your own setting — your scanners, your protocols, your case mix — remains the step no storefront can perform for you.
Nor does deployment end at go-live. Image-analysis models drift as scanners, protocols, and populations change, so the operational question after purchase is who runs the algorithmovigilance — the ongoing monitoring that catches a silently degrading model. Some platforms market continuous performance monitoring as a feature 3; confirm what is actually measured, how often, and who is accountable when a metric slips. And in every case the cleared apps are decision support: a human in the loop reads the study and owns the finding.
How to choose for your setting
Rather than a recommendation, a set of questions to hold up against any platform. The answers depend on your department, and they are yours to weigh.
- Does it carry the specific cleared apps I need? Start from the two or three indications that matter most in your reading room and check whether the platform lists a cleared app for each — rather than a roadmap item.
- Whose clearance, in my jurisdiction? An app cleared only under EU MDR is not authorized for US clinical use, and vice versa. Match clearance to where you practise 56.
- First-party or aggregator — and does that fit my risk posture? One accountable vendor versus a wider catalog with distributed accountability is a governance choice as much as a technical one.
- How does it touch my PACS/RIS and worklist? Integration depth determines whether findings reach the radiologist in workflow or in a separate window 46.
- Who monitors performance after deployment, and how will I know it drifted? Ask for the monitoring cadence and the escalation path, rather than the marketing line 3.
- What does the platform add to the clearance record? If the answer is "nothing," plan your own local validation before the tool touches a report 8.
For the wider growth and specialty picture behind these platforms, our FDA-cleared AI devices tracker follows the census itself; for the build-versus-buy question one layer down, see open vs closed models for hospital deployment.
Sources and method
This comparison draws its regulatory floor from the FDA's own AI-Enabled Medical Device List 1 and the ACR Data Science Institute's AI Central directory 2, its per-platform attributes from each vendor's official documentation 3456, and its context on what cleared imaging AI actually does from two peer-reviewed analyses of the FDA record 78. We include only platforms with a verifiable primary source, present attributes neutrally, and name no winner — inclusion here is not an AIMOCS endorsement of any product. Vendor-stated application counts and clearances are perishable; we revisit this page on a 180-day cycle and whenever the FDA updates its list or a platform changes what it carries. As of July 2026.