University Hospital Basel: AI finds its way into everyday hospital life
The University Hospital Basel (USB) uses artificial intelligence to support clinical processes. Patient data from various medical systems is automatically merged and processed using large language models (LLMs). Findings from these projects could be groundbreaking for the entire healthcare sector.
2024-12-09, 14:00
Doctors spend around half of their working time using IT systems, mainly for information searches and documentation. The University Hospital Basel is now using artificial intelligence in everyday practice to support staff with these tasks.
New era of precision medicine thanks to large language models
Until now, patient health data was stored in the specialized IT systems of the individual departments - from radiology to pathology to the laboratory - and mostly in the form of free text. With a clinical data warehouse, the USB initially created the technical basis for standardizing this fragmented data. With the use of LLMs, these AI systems can understand medical free text and process the relevant information. What used to take hours of manual compilation is now done in a matter of seconds. An internal study showed that the recommendations of the LLMs match the decisions of interdisciplinary tumor boards in 97% of cases. In prostate cancer treatment, for example, the system is already being evaluated by Dr. Wetterauer's team in order to display treatment options based on evidence-based guidelines.
Technical infrastructure and data protection
The USB uses a powerful GPU cluster that enables sensitive patient data to be processed completely locally. This is in line with the requirements of the new Swiss Data Protection Act. A specially developed governance structure ensures compliance with ethical and security requirements.
The new future of everyday practice with artificial intelligence
LLM technology is currently being tested in oncology, dermatology, psychosomatics, pathology, radiology and others. This follows on from the USB's previous successes in image recognition using convolutional neural networks (CNN) in radiology. The broad participation of different departments demonstrates the versatile application possibilities of the technology, from diagnostics to the optimization of administrative processes. Findings from these projects could be groundbreaking for the entire healthcare sector.
This press release was written with the help of artificial intelligence.
Media information center
