Yale researchers have demonstrated a machine learning tool for choosing between coronary imaging and stress testing in patients who present with suspected coronary artery disease.
In a study current in European Heart Journal, the team shows how the personalized-care tool consistently picked the optimal exam for achieving good outcomes in more than 2,100 patients.
Additionally, the tool avoided algorithmic bias by leveraging both arms of a major clinical trial to neutralize the skewing effect that real-world clinical decisions can have on study data.
“Our approach synthesizes the complex relationship between a large number of pre-randomization characteristics in creating and visualizing a comprehensive phenomap of patients, with an individualized assessment of the risk of adverse cardiovascular events with anatomical or functional testing for assessing chest pain,” the authors write.
The team recorded a significantly reduced risk of adverse cardiac events in patients whose exam choices matched those the AI tool would have recommended had it been there.
In Yale’s own coverage of the work, Evangelos Oikonomou, MD, DPhil, says the AI tool is technically sophisticated but practical for clinical settings.
“It relies on routinely captured patient characteristics and can be used by clinicians with a simple online calculator or can be incorporated in the electronic health record,” he says.
The study is posted in full for free.