How artificial intelligence detects invisible signs of heart failure

Researchers at Mount Sinai Hospital have invented a special artificial intelligence (AI)-based computer algorithm that can recognize subtle changes in electrocardiograms (also called ECGs or EKGs) to predict whether a patient has heart failure

"We proved that deep learning algorithms can identify blood pumping problems on both sides of the heart from ECG waveform data," said Dr.

The research led by Akhil Vaid is a doctor of medicine, a postdoctoral scholar engaged in the Glicksberg laboratory and a work by Girish N.

Heart failure, or congestive heart failure, occurs when the heart pumps less blood than the body normally needs, affecting approximately 6.

However, recent breakthroughs in artificial intelligence have shown that the electrocardiogram-a widely used electronic recording device-may be a fast and easily available alternative in these situations

"Although it is very attractive, the use of ECG to diagnose heart failure has traditionally been a challenge for doctors

Generally, an electrocardiogram consists of two steps

In this study, the researchers designed a computer to read a patient's electrocardiogram and data extracted from a written report that summarized the corresponding echocardiogram results of the same patient

Natural language processing programs help computers extract data from written reports

Dr.

Then, the computer read more than 700,000 ECG and echocardiogram reports obtained from 150,000 Mount Sinai medical system patients from 2003 to 2020

"One potential advantage of this study is that it involves the largest ECG collection in one of the world's most diverse patient populations," said Dr.

Preliminary results show that the algorithm can effectively predict which patients have healthy or very weak left ventricles

The algorithm predicts 94% of patients with healthy ejection fractions, and 87% of patients with ejection fractions lower than 40%

However, the algorithm is not as effective in predicting which patients' hearts will weaken slightly
.
In this case, the accuracy of the program in predicting patients with ejection fractions between 40% and 50% is 73%
.

Further results show that the algorithm has also learned to detect the weakness of the right valve from the electrocardiogram
.
In this case, the definition of frailty is a more descriptive term extracted from the echocardiogram report
.
The accuracy of the algorithm in predicting which patients have weaker right valve functions is 84%
.

"Our research results indicate that this algorithm may eventually help doctors correctly diagnose failure on both sides of the heart," said Dr.
Vaid
.

Finally, further analysis showed that the algorithm can effectively detect heart failure in all patients, regardless of race and gender
.

"Our findings indicate that this algorithm may be a useful tool to help clinicians fight heart failure suffered by various patients," Dr.
Glicksberg added
.
"We are carefully designing forward-looking trials to test their effectiveness in a more realistic environment
.
"

This research was supported by the National Institutes of Health (TR001433)
.

Article retrieval:

Vaid, A.
, et al.
, Using deep learning algorithms to simultaneously identify right and left ventricular dysfunction from the electrocardiogram, Journal of the American College of Cardiology: Cardiovascular Imaging, October 13, 2021, DOI: 10.
1016/j.
jcmg.
2021.
08 .
004.