How AI Helped Surgeons Remove an 11mm Brain Tumor Without Risking a Patient’s Sight

UCL and University College London Hospitals reported the first use of real-time AI assistance during live brain-tumor removal at London’s National Hospital for Neurology and Neurosurgery. The system helped surgeons navigate around blood vessels and nerves near the pituitary gland while removing a benign tumor from 48-year-old Rhys Hibbert. The surgical team remained fully responsible for every decision. 1

Who was the patient?

Hibbert, a customer-services manager and community volunteer from Bedfordshire, had an 11mm benign pituitary tumor. It was discovered after he collapsed while walking and had a seizure in December 2024. As the tumor affected the area around the optic nerves, it threatened his vision and could eventually have caused blindness if left untreated. 1

The tumor was successfully removed and his sight was protected. The supplied clinical report does not provide a detailed account of his longer-term recovery, so the available evidence supports the immediate surgical outcome without establishing broader recovery results. 1

Why pituitary-tumor surgery is so difficult

The pituitary gland is located at the skull base, close to the carotid arteries and the optic nerves. That makes the surgical margin extremely demanding: removing too little tissue can leave part of the tumor behind, while damaging nearby anatomy can cause severe complications, including bleeding, stroke or loss of vision. 14

The operation used an endoscopic approach. Rather than relying only on scans taken before surgery, the AI analyzed the live video feed from inside the operating field. That gave the team an additional way to identify structures as the procedure unfolded. 1

What the AI actually did

The system was designed as decision support, not as an autonomous surgeon. It processed the endoscopic footage and helped identify or highlight:

• critical blood vessels and nerves near the tumor;
• the surgical instruments being used; and
• interactions between instruments and tissue. 111

In practical terms, the AI acted as an additional visual aid, flagging high-risk areas and regions to avoid while the surgeons decided how to proceed. It did not operate independently, control the instruments or override the medical team. 1

That distinction matters. A successful first case shows that real-time assistance can be used in a live operation; it does not show that an AI system can safely perform neurosurgery on its own or that it improves outcomes for patients generally.

Who developed and performed the procedure?

Professor Hani Marcus, a UCL Queen Square Institute of Neurology professor and consultant neurosurgeon at the National Hospital for Neurology and Neurosurgery, performed the operation with Danyal Khan, a UCL PhD candidate and UCL/UCLH neurosurgical resident who led the work. 1

The technology came from a multidisciplinary team at the UCL Hawkes Institute, with Dr Sophia Bano identified as the technical lead. The wider Surgical Robot Vision group focuses on real-time analysis of anatomy, cameras and instruments, as well as the integration of surgical and imaging data. 2

How the system was trained

The AI was trained and evaluated using a large annotated collection of earlier endoscopic pituitary-surgery videos. The team described the collection as comprising hundreds of videos; UCL’s 2023 annual review says the system had analyzed more than 200 videos and reached, in 10 months, a level of experience that would take an individual surgeon years to acquire. 14

That training was aimed at helping the model recognize the steps of surgery, instruments and instrument–tissue interactions. The underlying research also includes detecting potential intraoperative causes of complications and identifying critical anatomy during key stages of pituitary surgery. 11

What hardware and funding supported it?

The system ran on NVIDIA’s Clara IGX platform, which is designed for real-time AI applications in medical-device environments. The clinical trial was funded by the National Institute for Health and Care Research and Google, with additional support from the NIHR UCLH Biomedical Research Centre, the Royal College of Surgeons, EPSRC and Wellcome. 1

The “ChatGPT for surgeons” ambition

The long-term vision is a real-time surgical co-pilot that can understand what is happening in an operation, recognize anatomy and surgical actions, and provide timely feedback. UCL researchers have described this direction using the shorthand “ChatGPT for surgeons.” 12

That phrase describes a future goal rather than the current capability demonstrated in this operation. The evidence provided supports a focused visual decision-support system for pituitary surgery; it does not establish that the tool has general conversational abilities or can autonomously plan and perform procedures.

What happens next?

The operation was conducted as part of an early clinical trial. It is an important demonstration of feasibility, but one patient cannot establish how well the system performs across different surgeons, hospitals, tumor locations or patient groups. UCL and its partners plan broader clinical testing, although the supplied sources do not specify the larger trial’s final size, sites or timetable. 1

UK Health Innovation Minister James Frith presented the case as an example of AI enabling new forms of care while emphasizing the need for appropriate safety safeguards. NIHR’s Mike Lewis also linked the work to the wider use of technology and innovation in healthcare. 1

The central result is therefore promising but limited: AI was brought into a live, high-risk brain operation as an extra set of eyes, while trained neurosurgeons remained in control. The next test is whether that assistance can produce reliable benefits across many more operations without compromising safety.