Researchers developed an artificial intelligence (AI) model that can accurately determine if a firefighter is about to experience a sudden cardiac event, which results in more on-duty fatalities than smoke and fire.
Researchers at the National Institute of Standards and Technology (NIST), in collaboration with the University of Rochester and Google, have successfully utilized machine learning to predict potentially fatal cardiac events in firefighters.
The use of AI has yielded promising results, demonstrating a significant ability to detect abnormal heart rhythms, a key cause of sudden cardiac death, which accounts for approximately 40% of on-duty fatalities among firefighters – more than fire or smoke inhalation.
The research findings, recently published in the Fire Safety Journal, revealed that the AI model was able to correctly identify around 97% of abnormal electrocardiogram (ECG) samples in a unique dataset collected from firefighters.
This technology, if implemented, could help save the lives of firefighters, who are at a disproportionately higher risk of sudden cardiac events compared to other first responders, researchers said.
A firefighter is twice as likely to experience a sudden cardiac death than a police officer, and four times more than other emergency responders. Sudden cardiac death claimed the lives of 36 on-duty firefighters in 2022, according to the National Fire Protection Association.
Limited ability to cool off in strenuous working environments, as well as pushing through these situations without realizing they might be at risk, are all factors contributing to these statistics.
“Year after year, sudden cardiac events are by far the number one killer of firefighters,” NIST researcher Chris Brown was quoted as saying in a press release. “Cardiac events also cause career-ending injuries and long-term disabilities.”
This alarming trend spurred the research team to tap into a unique dataset collected a decade earlier by the University of Rochester. The dataset contained 24 hours of ECG data from 112 firefighters, during both their on and off-duty hours.
Dillon Dzikowicz, co-author from Rochester, emphasized, “The firefighter data we collected is so unique. Having robust data is essential to move our work forward and protect firefighters.”
The NIST researchers developed the Heart Health Monitoring (H2M) model, leveraging machine learning and Rochester dataset. The model was trained to recognize and classify normal and abnormal heartbeats indicative of irregular heart rhythms such as atrial fibrillation or ventricular tachycardia.
The high accuracy of the H2M model, however, was not replicated when it was trained using non-firefighter ECG data, yielding a 40% error rate. This underscores the importance of using specific and relevant datasets for training AI models.
“Using the right dataset to train the AI model was critical,” commented NIST researcher Wai Cheong Tam.
The research team envisions the H2M model being used in portable heart monitors that firefighters could wear on-duty. Such a device would provide real-time alerts to potential cardiac irregularities, acting as an on-the-spot AI cardiologist.
Furthermore, the model's applications could be extended beyond firefighting, benefiting other high-risk groups and even the general public, if trained with the right datasets.
Tam noted, “This technology can save lives. It could benefit not only firefighters but other first responders and additional populations in the general public.”
Researchers at NIST are also working on an AI prediction model that could give firefighters an early warning of a flashover, a lethal fire phenomenon, and a machine learning algorithm to predict backdraft, another potentially lethal fire event.
More from Cybernews:
Subscribe to our newsletter