Surviving Blood Poisoning

Sepsis is one of the most common causes of death worldwide. An early warning system employing AI-supported sensors can save lives. How Fraunhofer researchers want to enable the detection of health risks.

A throbbing finger wound, a tooth infection, a urinary tract infection: Blood poisoning always starts with an inflammation. If pathogens enter the bloodstream, sepsis can develop in a heartbeat. The immune system releases a flood of inflammatory mediators that dilate and damage blood vessels. Fluids enter tissue, blood clotting goes out of control, the risk of thrombosis increases and organs fail. Only one in two people survive a septic shock. This disease is one of the most common causes of death worldwide.
 

Because every minute counts in sepsis cases, researchers at the Fraunhofer Institute for Computer Graphics Research IGD have been looking for a solution in the HealthView project to detect typical symptoms as early as possible. “Our project partner, Hypros GmbH in Stralsund, sells a device that uses optical sensors to detect critical clinical situations such as restlessness or the onset of delirium,” explains Mario Aehnelt, Deputy Site Manager at Fraunhofer IGD in Rostock. “To expand its functions, we have jointly evaluated AI-supported contactless sensors that can determine key vitals such as body temperature, respiratory rate and pulse rate,” adds project manager Gerald Bieber.

Sensors providing a three-dimensional image of people in bed are ideal. Algorithms can determine respiratory rate from the raising and lowering of the blanket by only millimeters. Pulse rate can also be measured indirectly: Every heartbeat causes a minimal change in blood flow to the facial skin. Pulse rate can be derived from this alone. “Body temperature, on the other hand,” says Bieber, “is a very good estimate based on the data from thermal imaging cameras and the daily temperature measurement by the caregiver.” The vitals are processed locally and are only provided in evaluated form, minimizing data protection concerns.

The feasibility study at Fraunhofer IGD is the basis for further development into a marketable system. This is tantalizing for startups looking for an innovative product, for example. In the next step, the research team wants to bring doctors on board to teach the AI the clinical interpretation of the parameters collected: How should the escalation levels be defined?

The Fraunhofer team is simultaneously examining further ideas for scaleup. “Continuous data acquisition and precise analysis without personnel expenses can be interesting not only for other disease processes, but also for the sensory monitoring of animals, in occupational health management, in professions where safety is a concern or in autonomous driving,” says Bieber. Further development of digital patient twins is also being accelerated. “Continuous data collection acquisition improves our knowledge of the boundary between health and illness,” emphasizes Mario Aehnelt. “This development is still in its infancy. But it is already clear that there are exciting opportunities here.”