Unverified. We have not been able to confirm the details of this account.
Reported The Amplifier Brilliant
Singapore hospital's AI flags brain bleeds on CT scans in about six minutes
Singapore General Hospital now uses an AI tool called iScreen to spot brain bleeds on head CT scans in about six minutes, so urgent cases are read first. A radiologist still checks every result. Builders: give triage AI one narrow job and keep a human sign-off on every report. Readers: a faster flag is good news, but it is fair to ask what happens to the scans the AI does not flag.
When an AI triage tool speeds up urgent reads, ask what share of real bleeds it misses and who checks the scans it leaves unflagged. A faster queue only helps if the cases it doesn't flag still get a timely human read.
A bleed in the brain is an emergency. The sooner a doctor sees it on a scan, the sooner the patient can be treated. At Singapore General Hospital, an AI tool now helps those scans jump the queue.
On 21 September 2026, the hospital announced iScreen with Synapxe, Singapore's national health tech agency, and health tech start-up Iota MedTech. iScreen reads head CT scans around the clock. It looks for signs of bleeding in the brain, which doctors call intracranial haemorrhage. When it spots one, it alerts the radiologists, the doctors who read scans, so the urgent case is looked at first.
It is the first head CT model on AimSG, Singapore's national AI platform for medical imaging. AimSG has been used in eight public healthcare institutions since 2023. Synapxe says this launch "opens the door" to bringing the tool to other public healthcare institutions.
The hospital does about 50 head CT scans a day. About one in ten shows an acute bleed. Urgent scans are usually reported in 30 minutes to an hour. Dr Robert Chen, who heads neuroradiology at the hospital and the National Neuroscience Institute, says iScreen's results are ready in about six minutes.
SGH says the tool has been part of its radiology work since January. Since then, reports on scans that show an acute bleed have come back within 30 minutes. Before, they took 30 to 60 minutes.
The machine does not get the last word. Every AI finding is checked by a radiologist before a final report goes out. The Straits Times reports that when iScreen flags a bleed, a text message goes to both the neuroradiologist and the patient's doctor. That nudges the scan ahead of everything else in the queue. Flagged scans are marked "ICH positive". The paper says other scans show no AI reading at all.
Why does a few minutes matter? "In haemorrhage care, that reduced reporting time can really make the difference between life and death, or between a full recovery and lasting disability," Chen said.
That is his view of why speed matters, and it is worth keeping in proportion. SGH has not published how many real bleeds iScreen misses. It has not published patient outcomes either. This is a hospital launch, not a peer-reviewed study. So the clear win on the record is speed: urgent scans read sooner, with a human still signing off.
The pattern is amplifier. iScreen has one narrow job: sort the scan queue so likely bleeds are read first. It does not replace the radiologist. It makes the radiologist's first look come sooner, which is exactly what it was built to do.
What to sit with
A narrow AI tool with a human sign-off can get urgent scans read sooner.
Worth arguing about
- Should hospitals publish miss rates before they roll triage AI out more widely?
- Is speed alone enough to call an AI tool a success in emergency care?
What’s at stake
- An AI tool now pushes likely brain bleeds to the front of the scan queue at a busy hospital.
- The hospital has shared speed gains, but not how many bleeds the tool misses.
What this gets right
- Urgent bleed reports now come back within 30 minutes, the hospital says.
- A radiologist checks every AI finding before a final report goes out.
What stays messy
- No missed-bleed or patient-outcome figures are public yet.
- The Straits Times says scans the tool does not flag show no AI reading at all.
Easy to miss
- This is a hospital launch, not a peer-reviewed study.
- We don't know yet how it performs at other hospitals on the same platform.
One thing to try What if every triage AI launch came with a public scorecard on the cases it missed, not just the time it saved?