AI Can Use Heartbeat Data to Calculate Overall Health
Recent studies have shown that artificial intelligence trained on subject data can analyse electrocardiogram records in order to estimate a patient’s sex and age. The team of researchers also concluded that with more time to develop, the AI could be used to form an overall reading of someone’s health and could be practically used by medical doctors to enhance their prescriptions and make problem identification more efficient.
This information is detailed in a scientific journal called “Circulation: Arrhythmia and Electrophysiology”, where it outlines how the researchers designed a machine learning tool to make predictions on patients based on their ECG data. An ECG, also known as an electrocardiogram, is a medical test used to record the electrical activity of the heart. It is simple and painless, and a useful tool in any professional’s toolbox.
The team of researchers that conducted the studies were from the Mayo Clinic College of Medicine and Science. For their AI, they made use of a type of algorithm known as a convolutional neural network which typically is developed to find parallels between inputs and outputs. They trained their machine learning algorithm on a total of 500,000 ECGs of different patients.

ECG data includes the specific patterns of a patient’s heartbeat. Source: theatlantic.com
In order to test the artificial intelligence programme, it was given ECG data on 275,000 patients and made to predict the age and sex of each patient. It managed to estimate the sex of the patient correctly 90% of the time whereas it only got the age of the patient right 72% of the time.
However, this may not mean that the AI just needs more training. Further investigation into why the algorithm couldn’t guess age as effectively found that it tended to overestimate age when confronted with patients with heart-related problems, such as past heart attacks or coronary artery disease, and underestimate when met with patients with no negative effects. In essence, the AI was in fact predicting the “physiologic age” of the patient, not their chronological age.
Being able to accurately and quickly assess the health of a patient could not only be useful for doctors and physicians, but also for the patients themselves, as they could be alerted much sooner to any health problems.
Thumbnail source: choosingwiselycanada.org