The Food and Drug Administration has cleared a new artificial intelligence platform from Aidoc that can scan a single abdominal CT and flag 14 different acute conditions at once, from appendicitis to spleen injury to obstructive kidney stones. The clearance, covering 11 newly authorized indications layered onto three the company already held, is one of the broadest single-workflow approvals the agency has granted for diagnostic imaging AI, and it lands at a moment when emergency departments across the country are struggling with imaging backlogs and radiologist shortages.
What the AI actually does
Aidoc’s system is built on CARE, a self-developed AI foundation model the company describes as healthcare’s first comprehensive foundation model to receive FDA clearance for this kind of broad triage task. Rather than running a separate narrow algorithm for each condition, CARE analyzes a single abdomen CT scan and simultaneously screens for a wide range of acute findings, including acute diverticulitis, abdominal-pelvic abscess, small and large bowel obstruction, intestinal ischemia, pneumatosis, kidney injury, liver injury, spleen injury and pelvic fracture. Those 11 new indications join three conditions Aidoc had previously secured clearance for: abdominal aortic aneurysm measurement, aortic dissection and intra-abdominal free air. The tool is delivered through Aidoc’s aiOS enterprise operating system, which is already installed in hospital imaging departments to route AI findings directly into radiologist worklists.
The numbers behind the clearance
In the pivotal study the FDA reviewed, the 11 newly cleared indications posted a mean sensitivity of 97%, reaching as high as 98.5% in certain settings, alongside a mean specificity of 98%, topping out at 99.7% for some conditions. Those figures matter because sensitivity determines how often the software correctly flags a genuine emergency rather than missing it, while specificity determines how often it avoids raising a false alarm on a normal scan. A tool that triages 14 conditions but generates a flood of false positives would simply add noise to an already overburdened radiology reading room; Aidoc’s data suggests the platform errs toward catching real disease without swamping clinicians with false flags.
Why hospitals are racing toward this kind of AI
The clearance arrives against a backdrop that radiology trade press has described as an inflection point for AI in medical imaging: thousands of health systems have already deployed some form of AI in clinical workflows, and by 2026 many hospitals treat these tools as standard digital infrastructure rather than experimental add-ons. The pressure driving that shift is structural. Radiology departments are contending with a persistent workforce shortage even as imaging volumes climb, and modern AI triage platforms have been credited with reducing scan turnaround times by 40% to 60% while improving radiologist productivity by 25% to 35% in some deployments. For a hospital emergency department, that can mean the difference between a ruptured appendix or bowel obstruction being flagged for a radiologist’s immediate attention versus sitting in a queue behind dozens of routine scans.
The case for AI triage — and its limits
Proponents argue that comprehensive triage tools like CARE represent a meaningful step beyond the single-disease algorithms that have dominated FDA’s AI device list for years, since a radiologist reading an abdominal CT is rarely looking for just one thing. A foundation model that screens for 14 conditions simultaneously more closely mirrors how a human reader actually works through a scan. But critics of the broader AI-in-radiology push caution that triage software is a workflow aid, not a diagnosis: FDA clearance for these tools does not certify that they replace a radiologist’s read, only that they can flag studies for prioritized review. Patient-safety experts have also flagged what they call the “AI diagnostic dilemma” — the risk that as more of these tools proliferate, hospitals may lean on automated flagging without adequate governance over how findings get escalated, verified and documented, potentially creating new failure points even as they solve old ones.
What’s next
Aidoc’s clearance is likely to intensify competition among radiology AI vendors racing to expand single-scan, multi-condition coverage, since a broader indication list gives hospitals more value from a single software license and a single integration into their picture archiving and communication systems. More than 1,400 AI-enabled medical devices already hold FDA marketing authorization across specialties, with radiology accounting for roughly three-quarters of that total, and comprehensive triage platforms like CARE are likely to keep expanding that count. The next test will be real-world performance data as more hospitals deploy the tool broadly — whether the sensitivity and specificity figures from Aidoc’s pivotal study hold up across the varied patient populations and scanner hardware found in everyday emergency departments, and whether the promised gains in turnaround time translate into measurably better patient outcomes rather than just faster paperwork.
Photo: TobiasRehbein / PIXABAY via Pixabay
