Science

AI promises to detect sepsis early and save lives

Sepsis, better known as blood poisoning, is a life-threatening condition. Occurs as a result of infection: pathogens such as bacteria, fungi or viruses enter the body, multiply there and attack the organs. However, the reaction of the body’s defenses is of decisive importance. Sepsis occurs only when pathogens overcome the defenses of the immune system, spread through the bloodstream, causing an overreaction of the immune system, causing the body to attack its own organs in the fight against infection.

An estimated 15,000 people develop sepsis every year in Switzerland, with a good third of cases resulting in death despite treatment. If left untreated, sepsis can lead to death within hours. Therefore, it is all the more important to recognize the first symptoms in time. Because the sooner the diagnosis is made and treatment is started, the higher the chances of survival.

Mortality risk reduced by 20%

To help healthcare professionals diagnose the symptoms of sepsis at an early stage, a team of researchers at Johns Hopkins University in the US has developed an early warning system that uses artificial intelligence. The researchers tested this system in a large-scale cohort study. It was attended by over 4,000 doctors from five hospitals and about 590,000 patients.

Key Finding: With an AI-based system, detection of sepsis symptoms can be completed faster—hours saved—than with existing methods, reducing the chance of patient death by 20%.

“This is the first time AI has been used at the bedside by thousands of healthcare providers and we are seeing lifesaving,” says Suchi Saria, founding director of the Malone Center for Health Engineering at Johns Hopkins University and lead author of the study. Thanks to this system, it will be possible to save thousands of patients with sepsis in the future. This approach will also be applied to other areas to improve other treatments.

Analysis of symptoms in relation to medical history

The system developed is called Targeted Real Time Early Warning System (TREWS). Johns Hopkins University said in a statement that it is reviewing medical records and medical records to identify patients at risk of life-threatening complications. The algorithm then compares the patient’s medical history with their symptoms.

and laboratory results. The system then tells the medical staff if the person is at risk of sepsis and to what extent. TREWS also offers appropriate treatment, such as antibiotics.

Success rate 82%

In total, AI correctly identified 82% of sepsis cases, the authors of the study write in the academic journal Nature Medicine. And in 38% of cases, the system issued an early warning, which the doctor could later confirm. In particularly severe cases, the early warning system detected the disease on average almost six hours earlier than with conventional methods.

Explainable AI

According to the press release, this result represents a marked improvement over machine learning and is intended to aid medical decision making by allowing doctors to understand why AI makes certain recommendations, according to Johns Hopkins University.

“This is a breakthrough in many ways. So far, most of these systems have been wrong far more often than they have been right,” said Albert Wu, a professor at Johns Hopkins University in Baltimore and co-author of the study. And such warning false positives can ultimately undermine the credibility of these systems.

Back to top button

Adblock Detected

Please consider supporting us by disabling your ad blocker.