Researchers at Gangnam Severance Hospital have developed an artificial intelligence model for predicting bacteremia, a pre-stage of sepsis, based on patients’ clinical data.
|Professor Song Young-goo|
Bacteremia is a condition where bacteria are present in the blood. The release of bacterial toxins into the bloodstream can lead to sepsis, which, in turn, can result in septic shock or death.
To develop the platform, the team, led by Professors Song Young-goo, Lee Kyung-hwa and Dong Jae-joon, and Selvas AI, a local AI company, analyzed more than 22,000 blood cultures from 13,402 patients diagnosed with bacteremia at the hospital. Afterward, the team used 1,260 data bacteremia set to train its AI platform and used another 210 bacteremia data to verify the efficacy of the AI model.
As a result, the team found it could obtain the most accurate predictive value when using 10 variables, including alkaline phosphatase enzyme levels.
“We developed an AI model that can detect bacteremia early by applying 10 clinical variables with high accuracy,” Professor Song said. “Unlike conventional AI models, which mainly analyze image data, the team’s platform analyzes actual clinical data of the patient’s vital signs such as body temperature, blood pressure and blood tests.”
The team expects that the platform will allow for a faster prognosis and real-time monitoring of acute infectious diseases such as sepsis, Song added.
The results of the trial were published in the October issue of the Journal of Clinical Medicine.
<© Korea Biomedical Review, All rights reserved.>