The Potential of AI-Based Clinical Text Mining to Improve Patient Safety: the Case of Implant Terms and Patient Journals
Marina Santini1, Oskar Jerdhaf2, Anette Karlsson2, Emma Eneling3, Magnus Stridsman3, Arne Jönsson4, and Peter Lundberg2,5
1Digital Health, Research Institutes of Sweden (RISE), Stockholm, Sweden, 2Radiation Physics, Linköping University, Linköping, Sweden, 3Unit for Technology Assessment, Testing and Innovation and Department of Health, Medicine and Caring Sciences, Linköping University, Linköping, Sweden, 4Department of Computer and Information Science, Linköping University, Linköping, Sweden, 5Center for medical Imaging and Visualization (CMIV), Linköping, Sweden
It is important for radiologists to know in advance if a patient has an implant, since MR-scanning is incompatible with some implants. At present, the unbiased process to ascertain whether a patient could be at risk is manual and not entirely reliable. We argue that this process can be enhanced and accelerated using AI-based clinical text-mining. We therefore investigated the automatic discovery of medical implant terms in electronic-medical-records (EMRs) written in Swedish using an AI-based text mining algorithm called BERT. BERT is a state-of-the-art language model trained using a deep learning algorithm based on transformers. Results are promising.
This abstract and the presentation materials are available to members only;
a login is required.