Most FDA-cleared AI imaging tools are built to catch one condition at a time, a stroke here, a fracture there. In January 2026, Israeli healthcare AI company Aidoc received what it and outlets like STAT News describe as the industry’s first comprehensive “foundation model” clinical AI triage clearance, a single workflow capable of flagging 14 different acute conditions from one abdominal CT scan.
What got cleared and how it works
The FDA clearance covers 11 newly authorized indications that, combined with three previously cleared ones, form one unified triage workflow. The newly cleared findings include appendicitis, acute diverticulitis, abdominal-pelvic abscess, small and large bowel obstruction, obstructive kidney stone, intestinal ischemia and pneumatosis, kidney injury, liver injury, spleen injury, and pelvic fracture, joining previously cleared indications for abdominal aortic measurement, aortic dissection, and intra-abdominal free air. The system is built on CARE, Aidoc’s internally developed foundation model, and is delivered through the company’s enterprise AI operating system, aiOS, which the company says has now analyzed more than 100 million patient cases.
The numbers behind the clearance
According to Medscape and Beckers Hospital Review, in the FDA-reviewed pivotal study the 11 newly cleared indications achieved a mean sensitivity of 97%, reaching as high as 98.5% in certain settings, and a mean specificity of 98%, up to 99.7%. Aidoc says the system also achieved roughly an order-of-magnitude reduction in false alerts compared with existing single-condition AI tools already on the market, a detail that matters because alert fatigue, radiologists ignoring or delaying review of AI flags after too many false positives, has been a persistent complaint about earlier-generation triage software.
Why a “foundation model” approach is a bigger deal than one more clearance
Aidoc already holds more than 31 FDA clearances deployed across roughly 2,000 hospitals, according to earlier reporting, but those were largely built and cleared one condition at a time. A single foundation model handling 14 conditions simultaneously means hospitals can potentially replace a patchwork of narrow point solutions with one system, simplifying both IT integration and radiologist workflow, since clinicians only need to learn and trust one triage interface rather than a dozen.
Perspectives and open questions
Radiology groups have generally welcomed the reduction in false alerts as a direct answer to burnout concerns that have dogged AI rollout in imaging departments, where large volumes of low-value alerts can add to, rather than reduce, radiologist workload. Skeptics note that pivotal-study sensitivity and specificity figures, however strong, still need to hold up as the tool is deployed across hospitals with different scanner hardware, patient populations, and imaging protocols than those used in FDA review, and that a single foundation model consolidating 14 conditions also concentrates risk: an error or blind spot in CARE’s underlying model could, in principle, affect triage accuracy across all 14 conditions at once rather than being isolated to one narrow algorithm.
Where this sits in the broader FDA AI landscape
Aidoc’s clearance lands against a backdrop of rapid overall growth in AI-cleared medical devices. As of March 30, 2026, the FDA had cleared 1,524 AI algorithms in total, with radiology accounting for 1,163 of them, or roughly 76% of all clearances, according to industry tracking of FDA data, even as radiology’s overall share of new clearances has been slowly declining as other specialties like cardiology and ophthalmology expand their own AI footprints. Competing radiology-AI vendor Viz.ai, by comparison, has racked up more than 50 separate FDA clearances across roughly 1,700 hospitals for its stroke and vascular detection tools, illustrating just how fragmented the single-condition clearance model has been up to now, and why Aidoc’s bundled 14-condition approach represents a distinct strategic bet rather than simply one more entry in a long list of narrow clearances.
What’s next
Aidoc has signaled this abdominal CT workflow is a template it intends to extend to other anatomic regions and modalities as its foundation-model strategy matures, rather than a one-off product. With FDA now clearing roughly 30 AI applications a month across specialties, according to industry tracking, foundation-model consolidation like Aidoc’s abdomen CT triage could become the direction other radiology AI vendors follow, moving the market away from single-condition tools toward broader multi-condition triage platforms.
How the workflow reaches radiologists in practice
Aidoc’s aiOS platform is designed to plug into a hospital’s existing picture archiving and communication system, or PACS, so that CARE’s triage flags appear directly inside the same viewer radiologists already use to read scans, rather than requiring a separate login or standalone application. That integration detail matters for adoption: AI tools that require radiologists to switch between systems tend to see lower real-world usage than those embedded directly into existing workflows, regardless of how strong their underlying accuracy metrics are. Reimbursement is a related open question; CMS has introduced specific billing codes for certain AI-assisted imaging analyses in recent years, but coverage and payment for foundation-model triage tools spanning 14 conditions at once remains less standardized than for single-condition AI, and hospitals will likely need to work through payer-specific documentation before the tool’s cost can be reliably offset by billing revenue.
