Hybrid machine learning for improved infection management in critically ill patients

More than 50% of Intensive Care Unit (ICU) patients have a serious infection and 70% of ICU patients are treated with at least one antibiotic. However, antibiotics are underdosed in up to 40% of them. This can lead to therapy failure, longer hospital stays and higher mortality. Moreover, this leads to long-term exposure to (possibly more powerful) antibiotics, which in turn promotes antibiotic resistance. The pharmacokinetics of antibiotics in these critically ill patients differ greatly from healthy individuals and are difficult to predict. It is therefore desirable to measure the antibiotic concentration in the blood. However, these blood tests are not routinely available, very labor intensive, and therefore expensive. HEROI2C aims to develop models that accurately predict the antibiotic concentration in the blood of a patient, using artificial intelligence, for the most commonly used antibiotics. Based on these predictions, dosing for the individual patient can be optimized by the doctors.

HEROI2C is collaboration between the ICU of UZ Ghent and IDLab of UGent-imec.

Project website

EOS special issue on health and technology with mBrain contribution from PreDiCT

The Eos special has three parts. The first is about prevention. Thanks to sensors, digital doubles that collect all our medical data and personalized medicines, we can detect and prevent more diseases than ever before. Part two focuses on therapy. New AI models allow doctors to design more efficient treatments, tailored to the patient. The final part focuses on care and how robots can lend a mechanical hand, and investigates the most pressing question of the moment: how can we overcome infectious diseases?

One of the contributions within this EOS special issue covers the HEROI2C project and focuses on our research on personalized antibiotics therapy in the intensive care.

EOS special issue on technology and health