Researchers at the Mayo Clinic set out to teach a computer how to identify heart disease using an EKG test in patients who do not yet have symptoms.
A type of heart disease, called asymptomatic left ventricular dysfunction, causes the heart to pump improperly without any obvious symptoms. It occurs in 3-6% of the population. With this type of heart disease, quality of life and expected years of living are reduced. However, if this disease is detected, it is treatable.
The problem is since this disease does not show any symptoms, patients may not know they are unwell and are thus unlikely to undergo testing. Currently, there is no simple screening test that can be done without discomfort or great expense to the patient. As a result, researchers at the Mayo Clinic in the USA carried out research to determine whether a simple EKG test could fit the bill for a such a test.
The EKG test involves placing 12 leads that stick onto the body at specific places around the heart. It is an inexpensive test and there is no needle prick necessary. The leads pick up the electrical impulses from the heart and form a line of varying heights, in which a wealth of information about the heart is contained.
It seems that doctors have not yet figured out the specific characteristics of the EKG test that would lead to a diagnosis. However, researchers hypothesized that computers may be able to pick up subtle changes that it contains, and they set out to create a form of artificial intelligence, similar to how computers have been taught to drive cars and translate languages. Their findings were recently published in Nature Medicine.
Researchers gave the computer data from over 625,000 patients who had had both an EKG test and an ultrasound of the heart. The computer was given the information on which patients had characteristics of asymptomatic left ventricular dysfunction. The form of artificial intelligence that was created was able to identify patients with this type of heart disease with 93% accuracy from the EKG test alone. This compares favorably with other types of screening tests doctor’s use, such as mammography for breast cancer with an accuracy of 85% and PAP test for cervical cancer with an accuracy of 70%.
Not only was the artificial intelligence that was created able to identify patients without symptoms, but it also unexpectedly had some accuracy in detecting individuals who were at future risk of developing the disease. This exciting technology enables a diagnosis to be made that extends beyond the capacity of the human skills of doctors.
Written by Nicola Cribb, VetMB DVSc Dip.ACVS
Reference: Attia Z, Kapa S, Lopez-Jimenez F, et al. Screening for cardiac contractile dysfunction using an artificial intelligence–enabled electrocardiogram. Nat Med. 2019;25(1):70–74. doi:10.1038/s41591-018-0240-2.