The National Research Program "Big Data" (NRP 75) has established the "ICU Cockpit" project, through which large amounts of intensive care data are used to develop procedures for early warning systems and therapeutic recommendations.
Patient safety in intensive care units could be significantly improved if false alarms are greatly reduced and critical complications such as epileptic seizures could be predicted. This is where the "ICU Cockpit" project of the National Research Programme "Big Data" (NRP 75) comes in: The large amounts of data from intensive care units are used to develop procedures for early warnings and therapy options.
A single critical patient treated in an intensive care or emergency ward generates up to 100 GB of data per day. The data is derived from patients’ monitoring, but also from examinations such as brain magnetic resonance tomographies, laboratory values, and biosensors. The flood of information often ends up not being used for the timely recognition of risk assessments or for rapid decision-making.
Conventional monitoring systems trigger around 700 alarms per patient a day, i.e. about one alarm every two minutes. A considerable proportion of these are false alarms. If the number of false alarms could be significantly reduced, the amount of data perceived as important to make an assessment would be much smaller, which would make it easier and quicker to identify critical situations and thus increase patient safety.
The Neurosurgical Intensive Care Unit at the University Hospital Zurich, the Swiss Federal Institute of Technology in Zurich (ETH Zurich) and IBM Research are working on this in the "ICU Cockpit" project. Project manager Emanuela Keller described the long-term goal: "With this project, we want to initiate a fundamental development in emergency and intensive care medicine, and thus significantly improve the way we work in everyday clinical practice".
For the project, multiple-sources data were systematically collected from more than 400 patients. Video recordings were also used. All data were anonymized before further processing. Patients in intensive care units are very vulnerable in various respects, so their data was particularly protected. From the data, the researchers developed procedures for three relevant targets:
1. Filtering out false alarms
2. Early detection of epileptic seizures
3. Early detection of secondary brain damage
The latter two procedures aim to identify risk constellations and warn of impending critical events, in order to strengthen the evidence for a prognosis. This allows earlier therapeutic intervention, which improves the quality of treatment.
Today, therapeutic decisions are often made empirically, based on the experience and knowledge of those involved. It would be desirable to underpin the decisions with data analyses available in real-time as well as the latest medical knowledge from other sources, e.g. globally harmonized databases. The project is gradually establishing the process to make this a reality.
The procedures are to be tested with further data sets and then directly implemented in the next study set up within the daily clinical routine of the University Hospital Zurich. The findings from the data analysis are to be presented visually and thus risk constellations automatically detected in patients’ intensive care units. In addition, work with IBM Research will continue to build processes in which video surveillance is used to detect epileptic seizures and other neurological disorders. These procedures are based on video recordings the data gathered, and the technologies created will help the research and monitoring of stroke patients with paralysis.