Scientists create algorithm that can detect abnormalities in heart rhythm

Scientists at Stanford University have developed an algorithm that can detect various types of heart rhythm abnormalities. The software would work as well as, and sometimes better than, trained cardiologists.

Stanford University has published the findings on its own website and in a scientific paper. The developed algorithm looks at the patient’s heart rhythm and can identify fourteen different types of abnormalities. In a study on the effectiveness of the algorithm, the results were compared with those of six trained cardiologists. In some cases, the software appeared to do as well as, and in some cases better than, the cardiologists.

To train the algorithm, a deep neural network was set up that was trained for seven months with a large dataset containing ECG data. The data was provided by the company iRhythm. The researchers provide more insight into the technical background of the software on a special website.

The algorithm can be especially useful for abnormalities that are not always visible on an ECG, or a heart film. In such a case, the patient must wear a mobile ECG machine that monitors the heart rhythm over a period of several weeks. Because the data then has to be checked manually, this is very labour-intensive.

Scientists at Stanford University previously came up with an algorithm that can detect skin cancer. There, too, they compared their method with specialist doctors, dermatologists, and found that the algorithm worked just as well.