🚨 This is a world’s first -
First brain surgery ever performed with a live AI assistant
A 48-year-old man walked into theatre at Queen Square with a pituitary tumour that was stealing his sight. He walked out able to see again.
👉🏻The difference this time:
AI was watching the live endoscopic feed in real time, monitoring pre-op scans and highlighting the optic nerves, blood vessels and other structures packed millimetres apart at the base of the skull.
One millimetre the wrong way and you get blindness, stroke or worse.
The surgeon stayed in full control. The AI just made a “visjble” difference.
That’s the first live AI-assisted neurosurgery of its kind.
Developed at
@uclh, trained on hundreds of previous operations, running on hardware built for real-time medical video.
The system runs on an
@nvidia Clara IGX platform designed to support real-time AI in medical device settings
🏥 Patient Rhys Hibbert volunteered to be first. His peripheral vision had collapsed to the point he needed walking sticks. Days later he was describing a “360-degree panoramic view” he hadn’t had in over a year.
How this could evolve?
This is still one model looking at one video stream.The logical next step isn’t a smarter single model. It’s a small team of specialised agents sitting alongside the surgeon:
- One watching anatomy and instruments
- One tracking heart rate, BP, oxygen, end-tidal CO2, temperature
- One watching blood loss and tissue colour change
- One watching EEG / depth of anaesthesia
- One modelling instrument–tissue interaction and force
- Others watching for early signs of swelling, bleeding or impending visual-pathway compromise
Ten Agentic AI streams could continuously predict of what is about to go right and what is about to go wrong, seconds before a human would notice.
The surgeon still decides. I think this is a game changer. The multi-agent monitoring layer is the obvious next trial.
Fantastic stuff
@JamesFrith @hani_marcus @SophiaBano @NVIDIAAI
Congratulations!!
$NVDA $DELL