Researchers at the Mayo Clinic trained a neural network on nearly 45,000 patients' paired ECGs and echocardiograms, then tested it on more than 52,000 more, asking whether the ECG alone, without an ultrasound, could flag a heart pumping at 35 percent capacity or less.
The network identified the weakened hearts with an accuracy measure, area under the curve, of 0.93, and people without a weak heart who screened positive were four times as likely to develop one later. A 10-second tracing was doing work that normally requires an ultrasound exam.
An abnormal AI score does not replace an echocardiogram; it flags who needs one sooner. A later randomized trial at Mayo Clinic found that giving doctors the AI result increased new diagnoses of a weak heart pump.
Read the original study
Attia et al., Nature Medicine, 2019 · doi.org/10.1038/s41591-018-0240-2
A century-old test, reading something new.
Source
- Attia, Z. I., Kapa, S., Lopez-Jimenez, F., McKie, P. M., Ladewig, D. J., Satam, G., Pellikka, P. A., Enriquez-Sarano, M., Noseworthy, P. A., Munger, T. M., Asirvatham, S. J., Scott, C. G., Carter, R. E., & Friedman, P. A. (2019). Screening for cardiac contractile dysfunction using an artificial intelligence-enabled electrocardiogram. Nature Medicine, 25(1), 70–74. https://doi.org/10.1038/s41591-018-0240-2