American researchers have developed artificial intelligence that helps doctors predict a patient’s risk of sepsis.
- Researchers at Ohio State University have developed an AI that can identify patients at risk of sepsis.
- This tool is able to identify if there is missing information to clarify the diagnosis and to request it from doctors.
- Testing showed that adding 8% of the recommended data improved the system’s sepsis prediction accuracy by 11%.
Sepsis, formerly known as septicemia, is an extreme response of the immune system to an infection, leading to organ dysfunction. Faced with this severe immune reaction, every minute counts to limit complications and fatal consequences.
Ohio State University researchers have developed artificial intelligence (AI) that helps doctors predict patients’ risks of sepsis. The tool, called SepsisLab, will be presented at the ACM KDD 2024 conference which will be held in Barcelona from August 25 to 29, 2024.
Sepsis: AI identifies missing data for reliable diagnosis
SepsisLab was developed to assess a patient’s risk of sepsis within 4 hours. However, it has a particularity compared to other artificial intelligences. It is able to determine if it is missing data (vital signs, demographic data, test results, etc.) and to ask doctors for them in order to improve the accuracy of its diagnosis.
“Experiments showed that adding 8% of the new recommended data from laboratory tests, vital signs and other variables reduced uncertainty in the model by 70%, contributing to an 11% improvement in sepsis risk accuracy.”specifies the American University press release.
Furthermore, while the tool was designed to provide a risk prediction quickly, it can also change it every hour after new patient data is added. This reduces diagnostic errors. “When a patient first comes in, there are many values missing, especially for lab tests”recalls the first author Changchang Yin.
Sepsis: joint work between doctors and AI
For the study’s lead author Ping Zhang, it is essential to develop a tool that allows efficient AI/human work. “The idea is that we need to involve AI in every intermediate step of decision-making by embracing the concept of AI in the human loop. We are not just developing a tool – we have also recruited doctors for the project. This is a true collaboration between computer scientists and clinicians to develop a human-centric system that puts the doctor in the driver’s seat.”he explains.
“The algorithm can select the most important variables and the doctor’s action reduces uncertainty”adds the scientist.
