Source/disclosure information
Issuer:
Disclosure: Avram reports receiving grants from the Quebec-en-Sante Foundation, the Montreal Heart Institute Research Center, and the University of Montreal, as well as private fees from Abbott, Biotronic, Boehringer Ingelheim, and Servier. ing. See this study for relevant financial disclosures of all other authors.
Important points:
- Artificial intelligence analysis of angiographic videos may help identify decreased left ventricular ejection fraction.
- AI may offer a ‘non-invasive alternative to left ventriculography’.
The researchers reported that deep neural networks demonstrated better discrimination of left ventricular ejection fraction depression using angiographic video compared with transthoracic echocardiography.
“Although the use of ventriculography has declined over time, it is still performed in up to 50% to 80% of angiograms, although it varies by institution, and remains the primary method for determining systolic function during coronary angiography. It is a cornerstone approach.Unexpected shortages of iodinated contrast agents have also recently limited the use of ventricular imaging.” Robert Abram, M.D., M.S., A cardiologist from the Department of Cardiology and Health at the University of California, San Francisco (UCSF) and a cardiac interventionist from the University of Montreal-Montreal Heart Institute and their colleagues wrote. “Video-based deep neural networks can learn subtle patterns from medical data to accomplish specific tasks beyond what doctors can accomplish with that data, without the need for additional cost or steps,” said Dr. It provides an opportunity to assess cardiac contractile function in real time from standard angiographic images.”
Image: Adobe Stock
AI to identify LVEF drops
Avram et al. trained and tested a video-based deep neural network called CathEF to identify LVEF decline and predict LVEF percentage on a large real-life patient dataset of clinical angiography videos from UCSF. and externally validated it on another dataset. University of Ottawa Heart Institute.
The results of this study are JAMA Cardiology.
A total of 4042 angiograms from 3679 adult patients were included in the analysis (mean age 64 years, 65% male).
On the UCSF dataset, CathEF identified a degraded LVEF with an area under the receiver operating characteristic curve of 0.911 (95% CI, 0.887-0.934). Diagnostic OR for decreased LVEF was 22.7 (95% CI, 14-37). According to the study, the mean absolute error in predicting LVEF percentage was 8.5% (95% CI, 8.1-9) compared with transthoracic echocardiography.
Avran et al. found that CathEF-predicted LVEF percentages differed by less than approximately 5% compared with transthoracic echocardiography in 38% of the dataset, but exceeded 15% in approximately 15.2% of the dataset. reported that a difference was observed.
Furthermore, according to this study, CathEF tended to overestimate low LVEF and underestimate high LVEF.
External Validation of AI for LVEF Prediction
In an external validation using the University of Ottawa Heart Institute dataset, the researchers found that CathEF reduced LVEF with an area under the receiver operating characteristic curve of 0.906 (95% CI, 0.881–0.931) and mean absolute error of LVEF prediction. reported that it identified The proportion was 7% (95% CI, 6.6–7.4).
In addition, studies show that the ability of CathEF to discriminate LVEF decline and predict LVEF percentage is affected by gender, BMI, low estimated glomerular filtration rate, presence of ACS, obstructive CAD and LV hypertrophy. was also consistent.
“The results of this cross-sectional study, to our knowledge, suggest for the first time that standard coronary angiography video may be used to provide an automated estimate of LVEF, thereby allowing clinicians to It provides information from coronary angiography alone that is not normally accessible, “usually requiring additional testing,” the researchers wrote. “Video-based deep neural networks represent a technological advance in coronary angiography that offers a new approach to assess LV systolic function that can be routinely applied during standard coronary angiography without the need for additional equipment or procedures.” It enables real-time dynamic assessment of cardiac function during coronary angiography and is a non-invasive alternative to left ventriculography for patients with additional risks and suboptimal contrast. We provide the means.”
