The artificial intelligence algorithm was 95% accurate in identifying patients with postpartum hemorrhage and detected 47% more cases of bleeding than standard methods. The research was published in the journal npj Digital Medicine.
Postpartum hemorrhage is common and can be life-threatening and sometimes requires removal of the uterus.
Scientists used the Flan-T5 model to analyze information on 131 thousand women who gave birth between 1998 and 2015. Their data was marked: Postpartum hemorrhage detected by doctors was identified with a special code.
The algorithm detected 47% more cases of bleeding compared to humans and showed an accuracy of 95%. The scientists’ findings will help future studies identify factors that may predict a high risk of postpartum hemorrhage.
In the future, scientists plan to test this approach, which could be used in real time to study other pregnancy complications.
Previous scientists to create A factor that increases the relative risk of vascular dementia by 78%.