A team of Danish researchers has developed a new artificial intelligence (AI) framework to deal with stroke numbers that are unrecognizable to human emergency callers. The framework outperformed emergency call handlers in recognizing stroke in men and women and all age groups studied, demonstrating its potential as an auxiliary tool for early and accurate stroke identification in the future.
Presented today at the European Stroke Organizations Congress (ESOC) 2023, the retrospective study uses data from the Danish Stroke Registry and over 1.5 million calls to the Copenhagen Emergency Medical Service between 2015 and 2020. It was drawn from the set, which included more than 7,000 stroke-related cases. call. The researchers used this data to train an AI framework to first transcribe phone calls and then predict stroke risk based on the transcribed text.
Results evaluated in calls from 2021 reveal that the AI framework performed more effectively than emergency call handlers in identifying stroke cases. The AI framework achieved a recall (sensitivity) of 63.0%, a precision (positive predictive value) of 24.9%, and an F1 score of 35.7. In contrast, emergency call handlers had a recall of 52.7% and a precision of 17.1%, yielding an F1 score of 25.8.
Dr. Jonathan Wenstrup of Copenhagen University Hospital, one of the study’s first authors, said: “As one of the first points of contact for patients seeking medical assistance, emergency call handlers can help facilitate early and accurate stroke recognition. plays an important role in,” he said. Many stroke cases may go undetected at this stage, leading to delayed treatment and potentially life-threatening consequences for patients. ”
Across Europe, stroke is the second leading cause of death and the leading cause of adult disability, affecting more than one million people each year. With a growing and aging population, the number of people suffering from stroke in the European Union (EU) is projected to increase by 27% between 2017 and 2047. Nonetheless, many strokes can be prevented, and early treatment greatly improves the chances of a favorable outcome.
“The introduction of this new cost-effective support tool will enhance stroke identification by call handlers, ensure more patients receive appropriate and timely care, and ultimately improve patient outcomes.” We can do that,” added Dr. Wenstrup.
“As with other new tools, further research and development is needed to improve the accuracy of the framework and extend its functionality. “Non-word speech, such as slurred speech, will be incorporated into the training data. It is already clear that it has the ability to completely transform diagnosis and care.”
Courtesy of the European Stroke Agency Conference
