Cardiologists in the US say they have developed an AI program that can calculate LVEF using regular angiogram videos, potentially bypassing the need for invasive procedures.

Named CathEF, the tool remains in early development but has already achieved accuracy rates above 90% on two test data sets, according to its creators.

Based on a type of algorithm called a deep neural network, the program was trained using patient data from 4042 adult angiograms with corresponding transthoracic echocardiogram (TTE) LVEF from 3697 patients treated at University Of California, San Francisco (UCSF) hospitals.

From this, the AI learned to distinguish reduced LVEF (≤40%) and to predict continuous LVEF percentage from standard angiogram videos of the left coronary artery, said lead developer Associate Professor Geoff Tison of UCSF.

“CathEF offers a novel approach that leverages data that is routinely collected during every angiogram to provide information that is not currently available to clinicians during angiography,” he said.

“This is effectively expanding the utility of medical data with AI and provides real-time LVEF information that informs clinical decision-making.”

Photo: CathEF

In a study published this month, the team reported outcomes from testing the data on a further 813 patients at UCSF, plus another 776 from Montreal Heart Institute in Canada.

These tests found the tool discriminated reduced LVEF with 91.1% and 90.6% accuracy respectively using AUROC – “good discrimination”, the authors said.

Mean absolute error of predicted continuous LVEF was 8.5% in the UCSF group and 7% in the Montreal group, they reported.

And pleasingly, the algorithm had similar performance regardless of sex, BMI, ACS, presences of obstructive coronary artery disease and presence and severity of LVH, the researchers said.

Nevertheless, when the program got it wrong, it tended to get it very wrong, the authors warned. Although AI-predicted continuous LVEF differed 5% or less compared with TTE LVEF in greater than a third of test studies, differences greater than 15% were observed in 124 of the 813 cases.

This suggested “caution is warranted at extremes of LVEF”, they wrote in JAMA Cardiology (link here).

But while further research is needed to improve the accuracy and reduce the variability of programs like CathEF, the technology has great potential, the team said.

“This work demonstrates that AI technology has the potential to reduce the need for invasive testing and improve the diagnostic capabilities of cardiologists, ultimately improving patient outcomes and quality of life,” said first author Dr Robert Avram of Montreal Heart Institute.

Further research was now underway to test the program at point-of-care and determine its impact on clinical workflow, the interventional cardiologist said.

He added that to this end, a multi-centre prospective validation study in patients with ACS was underway to compare the performance of CathEF and the left ventriculogram with TTEs performed within 7 days of ACS.