Hallym-Uppsala: Artificial intelligence in emergency care
Johan Sundström, Professor at the Department of Medical Sciences, will present his work at the Hallym-Uppsala symposium on 13 September, where the theme is “AI in health care”. Photo: Mikael Wallerstedt.
“The emergency department is an environment where I think AI can become extremely valuable. Doctors in emergency settings might have two minutes to check a patient’s medical record and prepare before the ambulance arrives. This is a very limited time for a human, but for an AI, two minutes is a long time. We believe that AI will contribute to saving lives in the emergency department in a near future.”
This says professor Johan Sundström who is studying the possibilities for using artificial intelligence to enhance safety, efficacy and cost-effectiveness in emergency departments. He will present his work at the Hallym-Uppsala symposium on 13 September, where the theme is “AI in health care”.
“The basis for our emergency care project is a very large amount of health data from 6 million emergency care visits, from multiple hospitals. We are training our AI models to provide decision support for emergency care staff, which has the benefit of having a clear starting point, that is, when the patient arrives,” Sundström explains.
“In these data, we have half a million digital ECGs that we have used to train an AI to detect myocardial infarctions. With this amount of data, the AI became very good at finding myocardial infarctions, and discriminating between large and small ones.”
For the identification of a myocardial infarction, it became apparent that the AI used patterns that are not used by humans for this purpose. So in a sense, the AI is now teaching doctors how to read ECGs.
“We are now investigating if we can retrain the ECG model to detect other conditions, for instance blood clots in the lung arteries or a ruptured aorta, that humans currently don’t know how to find in an ECG. These are dangerous conditions that are very important to identify in the emergency department,” Sundström says.
Johan Sundström thinks that the theme of the symposium is very relevant and timely. Other talks will present topics such as patient monitoring using AI, digital smart hospitals that apply AI-powered modules and the role of AI for optimising drug dosing.
“I’m really looking forward to the symposium, to learn from colleagues from near and far, and to be inspired by all the clever people working in this cross-disciplinary and lively research field.”
Kerstin Henriksson