The research project HEART explores a non-invasive method for monitoring the body’s fluid requirements based on ECG signals. AI-supported analysis of large datasets supports precision medicine and offers significant benefits for patients.
The research project HEART investigates a completely non-invasive method for accurately determining the body’s fluid requirements based on ECG signals. The use of artificial intelligence to analyze large retrospective datasets also represents a step toward precision medicine, where treatment decisions are made based on individual criteria.
In healthy individuals, fluid balance is regulated autonomously, but this mechanism fails during lengthy surgical procedures or under extreme stress. In long surgeries with blood loss or excessive evaporation, the patient’s fluid requirement must be determined precisely via advanced hemodynamic monitoring, as significant deviations from the optimal range can cause serious complications. All currently available methods for advanced monitoring are invasive (i.e., they require puncturing blood vessels) and thus prone to complications. There is currently no precise, reliable, easy-to-use non-invasive procedure for advanced monitoring.
The basis for the targeted non-invasive method is the assumption that changes in fluid balance also affect the heart’s conduction system and that these changes must be reflected in the ECG. As part of HEART, large volumes of retrospective data from the Kepler University Hospital Linz and from suitable public databases are analyzed to prove a correlation between ECG signal changes and fluid administration. State-of-the-art deep learning methods (e.g., CNN, LSTM, Transformer) are used to analyze the ECG segments, alongside more interpretable approaches based on established manually extracted ECG features using traditional machine learning. The insights gained may then be validated in a potential follow-up project using prospective data.
In selecting data, HEART will pay particular attention to a balanced gender distribution and a representative age distribution in order to prevent data bias and systematic prediction errors. The resulting analysis results will be reviewed for their applicability across different subpopulations.
The implementation of this novel and disruptive method would offer a much-needed, reliable, cost-efficient, and user-friendly patient monitoring solution with several advantages of a non-invasive approach: for anesthesiologists, an easy-to-use but reliable first stage of advanced hemodynamic monitoring; for nursing staff, a time-saving application with minimal training required; for patients, a reduced risk and potentially shorter hospital stays; for hospital operators, a cost-effective solution. An extended benefit of non-invasive hemodynamic monitoring could be found in extended (institutional) care or even leisure applications. Through low-threshold ECG measurements (e.g., wearables), fluid requirements of elderly individuals or extreme athletes could be monitored.
The research activities in HEART are being carried out in close coordination with the University Clinic for Anesthesiology and Operative Intensive Care Medicine at the Medical Faculty of JKU Linz, as well as the Clinical Department of General Anesthesia and Intensive Care Medicine at MedUni Vienna.
This project is funded by the FFG.
Head of Unit Medical Informatics
