A groundbreaking AI model developed by Johns Hopkins University researchers can detect previously unseen signals in ECG tests, predicting potentially deadly post-surgical complications with unprecedented accuracy.
A newly developed artificial intelligence model is set to revolutionize how surgeons predict and manage post-surgical complications, significantly outperforming traditional risk scores currently used by doctors. This innovative breakthrough comes from researchers at Johns Hopkins University, who have utilized AI to uncover previously undetected signals in routine electrocardiogram (ECG) tests.
By tapping into more intricate physiological data within ECG results, the AI model promises to improve decision-making and risk calculations for both doctors and patients.
“We demonstrate that a basic electrocardiogram contains important prognostic information not identifiable by the naked eye,” senior author Robert D. Stevens, the chief of the Division of Informatics, Integration, and Innovation at Johns Hopkins Medicine, said in a news release. “We can only extract it with machine learning techniques.”
Published today in the British Journal of Anaesthesia, the study addresses a critical need in the medical community.
Traditional risk scores, used to identify patients likely to face complications from surgery, hold an accuracy rate of just 60%.
Approximately 37,000 patients’ preoperative ECG data were analyzed to train the AI models, with one model focusing solely on ECG data and another “fusion” model integrating ECG results with patient medical records.
The results were compelling. The ECG-only model already surpassed existing risk assessments, but the fusion model went further, achieving an impressive 85% accuracy in predicting post-surgical complications such as heart attacks, strokes or death within 30 days of surgery.
“Surprising that we can take this routine diagnostic, this 10 seconds worth of data and predict really well if someone will die after surgery,” added lead author Carl Harris, a doctoral student in biomedical engineering. “We have a really meaningful finding that can improve the assessment of surgical risk.”
Stevens elaborated on the broader implications of these findings, highlighting the extensive range of physiological information that an ECG can capture, including data on inflammation, the endocrine system, metabolism, fluids and electrolytes.
“If we could get a really big dataset of ECG results, and analyze it with deep learning, we reasoned we could get valuable information not currently available to clinicians,” Stevens added.
The team also created a method to identify which ECG features could be linked to a heart attack or stroke following surgery.
“You can imagine if you’re undergoing major surgery, instead of just having your ECG put in your records where no one will look at it, it’s run thru a model and you get a risk assessment and can talk with your doctor about the risks and benefits of surgery,” added Stevens. “It’s a transformative step forward in how we assess risk for patients.”
The new AI model not only provides a significant leap in predicting surgical risks but also offers insights that could open further research avenues.
The team is keen to test the model on larger datasets and in prospective studies involving patients about to undergo surgery.
The team is also exploring other potential valuable data points that AI might extract from ECGs in the future.
Source: Johns Hopkins University