Academic case study — not a medical device
How an AI triage system changes WHEN a radiologist sees a case — never WHO is responsible
This demo illustrates, in a simplified and fictional way, the concept behind Aidoc: an AI that analyzes scans in the background and reorders the radiologist's reading list to surface urgent cases — without ever replacing their clinical judgment.
Launch the interactive demo →The problem: a queue with no clinical priority
In a radiology department, exams pile up in the order they arrive (FIFO), with no signal of clinical urgency. A critical case — stroke, pulmonary embolism, hemorrhage — can end up waiting behind several routine exams simply because it arrived after them.
What Aidoc does: re-prioritize, never diagnose
A few minutes after a scan is acquired, the AI analyzes it in the background, directly inside the PACS (the software the hospital already uses). If it detects a sign of severity, the case moves to the top of the radiologist's worklist. It only reorders the queue — it never makes a diagnosis.
What doesn't change: medical accountability
The radiologist still reads and interprets every case themselves. They sign the final report and remain solely legally responsible. The AI changes WHEN a case is seen — never WHO is responsible for the diagnosis.
A real challenge: generalizability
On its 2026 foundation model (14 pathologies), Aidoc reports 97% sensitivity and 98% specificity in internal validation. These figures drop in external validation — outside the training hospitals — where specificity can fall by up to 24 points. This is a genuine research challenge for this type of system, illustrated here in a simplified way.