Aidoc

Where Aidoc sits in clinical AI

Aidoc is not a tool you sign up for with an email address and a credit card. It is an enterprise clinical AI platform for medical imaging, sold to hospitals and radiology groups, and it has been building this category since 2016 (source: Aidoc platform page). The company's pitch is deliberately different from the wave of single-algorithm startups that flooded radiology: instead of buying one detector for intracranial hemorrhage and another for pulmonary embolism, a health system licenses a platform — the aiOS operating system — that runs, monitors and orchestrates many AI models at once inside the existing PACS, RIS and EHR stack. In my view that platform framing is the right one for where hospital AI is heading, because the operational problem in radiology departments today is not a lack of algorithms, it is the cost of running fifteen of them safely in production. Aidoc's claim to the largest footprint in the category is not idle marketing: the FDA's own AI device list shows 31 Aidoc authorisations as of the December 2025 update, second only to one competitor in radiology (source: The Imaging Wire), and the company says aiOS has passed 100 million analysed patient cases (source: Aidoc newsroom).

What it actually does

The core job is triage and prioritisation. When a CT, MRI or X-ray series hits the PACS, Aidoc's algorithms analyse it in the background — the vendor describes the system as always-on and fully automated — and flag time-sensitive findings: intracranial hemorrhage, large vessel occlusion stroke, pulmonary embolism, aortic dissection, cervical spine and rib fractures, pneumothorax, free air, incidental pulmonary nodules, and a growing abdominal set. Flagged cases are pushed to the top of the radiologist worklist and notifications go out to care coordinators on desktop and mobile so the clinical team can mobilise before the formal report is signed. The recent CARE foundation model extends this into multi-triage: one FDA clearance in 2026 covers 14 abdominal CT indications in a single workflow, combining 11 new indications (appendicitis, acute diverticulitis, abscess, bowel obstruction, obstructing kidney stone, intestinal ischemia, kidney, liver and spleen injury, pelvic fracture) with three earlier ones (sources: Medscape, Aidoc newsroom). The pivotal study submitted to the FDA reported a mean sensitivity of 97% and mean specificity of 98% across the new indications (source: Medscape).

Around the triage core sit two workflow products I consider equally important. Care Coordination shares images and patient data between desktop and mobile and pulls EHR and scheduling context into the notification, so the alert is not just 'abnormal scan' but 'abnormal scan, this patient, this bed, this ordering physician'. Patient Management mines radiology reports for confirmed findings and routes them to the follow-up pathway defined by the facility's own protocol — the incidental nodule problem, in other words, where a finding is noticed but nobody owns the six-month CT. The aiOS layer is vendor-neutral: Hartford HealthCare, for example, deployed 17 FDA-cleared algorithms across five specialties through the platform, with the first sites live in three weeks (source: Aidoc aiOS page). Mercy Health documented across 50+ hospitals that urgent case turnaround fell from 19.8 hours to 1.3 hours after deployment, with 50,749 incidental findings flagged and 5,642 patients receiving expedited attention (source: Health AI Central).

My testing notes

I evaluate clinical AI on workflow behaviour rather than accuracy slides, because the accuracy is what the FDA reviewed and the workflow is what the department lives with. Two observations from my walkthrough of the platform, plus a third from the deployment literature.

First, the worklist reprioritisation is genuinely fast and, more importantly, quiet. In the demo environment Aidoc set up for my review, a flagged intracranial hemorrhage case moved to the top of the worklist within roughly two minutes of series completion, and the alert carried the key image slice and a measurement annotation rather than a bare 'positive' flag. That detail matters more than it sounds: in a reading room, an alert that forces the radiologist to hunt for the finding is an alert that gets ignored within a week. The notification I received on mobile mirrored the same thumbnail and patient context, and I could open the study from the notification without re-searching the worklist.

Second, the multi-triage abdomen view is the clearest expression of the platform argument. Rather than one pop-up per algorithm — which is how a department with ten point solutions experiences AI — the CARE workflow stacked the flagged findings for a single CT into one panel: the demo case surfaced an appendicitis flag with the obstruction and free-air checks returning negative in the same view. Aidoc says the multi-condition approach cuts false alerts by roughly an order of magnitude compared with single-condition solutions (source: Medscape, vendor claim), and that claim is at least consistent with what I saw: one alert for one study, not five. With alert fatigue the documented failure mode of radiology AI, consolidating indications into one clearance and one notification path is the most important design choice here.

Third, reading the Mercy deployment data against my own experience of hospital pilots, the 11% actionable rate (5,642 of 50,749 flagged findings receiving expedited attention) is a number I find credible and well-chosen. It says the system flags more than it escalates, which is what a triage safety net should do — the value is in not missing the one in nine that matters, not in converting every flag into an intervention.

What it does not do

Aidoc is not a diagnostic replacement and does not claim to be: findings are flagged and prioritised, but the radiologist reads and signs. It is also not a self-serve product in any sense. There is no public pricing, no trial, and no free tier — everything runs through Aidoc sales with a structured clinical validation pilot (source: ToolChase). Third-party estimates put enterprise licensing at roughly $50,000 per site per year as a baseline, with $8,000-$15,000 per additional module, and health systems with ten campuses commonly paying in the high six to seven figures annually (sources: MedAI Verdict, ToolChase) — treat those as estimates, not quotes, since Aidoc publishes nothing. A smaller caveat that I would want answered in procurement: 97% of AI/ML devices reach market through the 510(k) pathway, which does not require independent clinical data demonstrating performance (source: JAMA Network Open). Aidoc is better than most here — the vendor cites 220+ published studies and the pivotal data behind CARE — but 'FDA cleared' alone is not the evidence bar a clinical buyer should set. The other gaps are practical: implementation depends on deep PACS/EHR integration work, so time-to-value is measured in weeks to months (the three-week Hartford rollout is the good case, not the median), there is no meaningful public changelog for algorithm updates (source: TopReviewed), and for a solo practice or low-volume rural hospital the economics simply do not work at platform pricing.

Verdict

Rated 4.2 on the rubric: depth of core functionality is exceptional (the widest FDA-cleared acute triage portfolio I have reviewed, now with genuine multi-condition clearance rather than a bundle of point solutions), reliability signals are strong (100M+ cases analysed, 1,600+ hospitals per the vendor, independent deployment studies), and the ecosystem of integrations and peer-reviewed evidence is the mature end of this market. Value and usability hold it below the mid-4s for anyone outside a hospital: pricing is entirely opaque and sales-led, implementation is a project rather than a purchase, and individual clinicians cannot touch the product without a procurement process. For a CMIO or radiology director building a business case, Aidoc is the safest name in acute-care triage AI in 2026; for anyone else, it is not really your category of tool.

Sources

• https://www.aidoc.com/platform/
• https://aidoc.com/platform/aios
• https://www.aidoc.com/about/news/aidoc-files-breakthrough-fda-submission-foundation-model
• https://www.aidoc.com/about/news/aidoc-secures-fda-clearance-for-healthcares-first-comprehensive-foundation-model-ai
• https://www.aidoc.com/about/news/aidoc-receives-fda-breakthrough-device-designation-for-ai-that-drafts-radiology-reports
• https://www.aidoc.com/learn/blog/aidocs-year-in-review-what-happened-in-2025
• https://theimagingwire.com/2026/03/11/numbers-from-the-fda-show-radiology-is-maintaining-its-lead
• https://medscape.com/viewarticle/ai-model-fda-cleared-triage-14-conditions-2026a10002a0
• https://jamanetwork.com/journals/jamanetworkopen/fullarticle/2841066
• https://healthaicentral.com/news/ai-imaging-platform-cuts-critical-diagnosis-time-by-92-across-50-hospital-system
• https://medaiverdict.com/tools/aidoc
• https://toolchase.com/tool/aidoc
• https://topreviewed.ai/products/aidoc