Neurosurgeons from the Faculty of Medicine of Comenius University Bratislava and the University Hospital Bratislava performed Central Europe’s first AI-assisted brain surgery
Doctors from the Department of Neurosurgery of the Faculty of Medicine, Comenius University Bratislava (CU), and the University Hospital Bratislava (UNB), based at Kramáre, have become the first in Central Europe to remove a brain tumour with the assistance of artificial intelligence. The procedure was carried out in close collaboration with a Spanish neurosurgeon. The outcome is encouraging: postoperative MRI showed no residual tumour, and the patient has already been discharged to home care.
In primary brain tumours, so-called gliomas, one of the main challenges is that tumour tissue blends with healthy brain tissue. While the tumour is visible on MRI, it is often difficult to distinguish it reliably from surrounding tissue during surgery. “The more tumour we can safely remove, the longer the patient survives,” explains Head of the Department of Neurosurgery Andrej Šteňo. At the same time, the neurosurgeon must protect functional centres responsible for critical processes such as speech and comprehension, which is why such procedures are often performed with the patient awake.
The team at Kramáre has long-standing experience with intraoperative ultrasound, one of the most accessible and effective methods for imaging tumour tissue during surgery. Its interpretation, however, is complicated by artefacts - reflections of the ultrasound signal that can obscure the true anatomical situation and make it harder to distinguish between residual tumour and healthy brain. This is where artificial intelligence comes in, helping the surgeon identify where the tumour actually continues and where healthy tissue begins.
Ultrasound and AI: partners in the operating theatre
The AI model was developed by Spanish neurosurgeon and programmer Santiago Cepeda from Hospital Río Hortega in Valladolid. It is based on deep learning and computer vision. Dr Cepeda trained the model on more than 1,400 ultrasound images of brain tumours. The system has learned to distinguish tumour tissue, the peritumoural zone, and selected anatomical structures, while evaluating images in real time during surgery. “The aim of the project is to validate this model in other external centres. Validation in a different clinical environment is a key hurdle these methods must overcome,” he adds.
In the operating theatre, the ultrasound device is connected to a high-performance computer running the model, which displays colour-coded regions representing the tumour, its surroundings, and the resection cavity. “I would perform the operation the same way even without it. But if I overlook something and the model flags it, I will certainly take another look,” says Associate Professor Šteňo. In his view, AI functions as an additional expert opinion in the team, an “external brain”, rather than a substitute for human decision-making. Responsibility for the patient always remains with the surgeon.
The Bratislava neurosurgical team was not selected for the project by chance. Dr Cepeda had been seeking centres with strong expertise in intraoperative ultrasonography and repeatedly encountered Kramáre and Associate Professor Šteňo in the literature. In addition to Bratislava, he approached the renowned Istituto Carlo Besta in Milan, forming a three-member European consortium.
In Bratislava, however, the project progressed at an exceptional pace. While large European centres are often slowed down by complex administrative procedures, the Slovak team demonstrated a high level of readiness. “The preparedness of our Slovak colleagues was very favourable. Approval and all necessary steps were completed within a few months, largely thanks to strong interest,” says Cepeda, noting that while ethics approval processes are still ongoing in Milan, doctors in Bratislava have already successfully performed AI-assisted surgery.
The research and publication outputs of the Kramáre team have also recently contributed to changes in the latest recommendations for the surgical treatment of gliomas, which now reflect their data from procedures using intraoperative ultrasound. “Being approached by a leading expert such as Dr Cepeda is already the second major recognition of our work in a short period of time,” adds Associate Professor Šteňo.
No residual tumour, strong prospects
The first AI-assisted brain surgery at Kramáre was performed on a patient with a glioma, where maximal tumour removal while preserving functional centres is critical. Postoperative MRI showed no evident residual tumour, and the patient was discharged to home care within a few days.
The model developed in Valladolid is already used routinely there. In collaboration with Bratislava, further refinement and testing on a larger cohort of patients are planned. “If the model can be further trained and very high reliability can be achieved, it could shorten surgery times while increasing safety,” says Andrej Šteňo.
Artificial intelligence in medicine: support rather than replacement
According to both experts, the project demonstrates how modern technology should be used in medicine: as a tool that enhances decision-making rather than a replacement for human expertise. Associate Professor Šteňo also stresses the need for clear rules governing the use of AI, to prevent clinicians - especially those in training - from delegating their own reasoning to technology while still developing their clinical judgement and scientific writing skills.
For Santiago Cepeda, the greatest satisfaction lies in technological development, but he emphasises that its true value will be realised only when patients benefit directly in the near future - through shorter time in the operating theatre, safer tumour removal, and, above all, improved prognosis in their serious disease.
Eva Kopecká