The deep learning algorithm is able to analyze hours worth of data from some wearable heart monitors in order to detect potentially life-threatening arrhythmias. Not only does it outperform trained cardiologists, say the researchers, but it also has the advantage of being able to sort through data from patients located in remote conditions who have no access to cardiologists.
One of the key takeaways from this research, says Awni Hannun, a graduate student at Stanford University and co-lead author of a paper on the algorithm, is not just the abnormality detection, but that “we do it with high accuracy across a large number of different types of abnormalities. This is definitely something that you won’t find to this level of accuracy anywhere else.”
Typically, patients with suspected arrhythmias may need to wear a wearable electrocardiogram monitor continuously for weeks at a time in order for it to collect enough data to include any potential problems. The resulting data then needs to be inspected second by second to detect and then differentiate between harmless heartbeat irregularities and those that are problematic.
“The differences in the heartbeat signal can be very subtle but have massive impact in how you choose to tackle these detections,” says Pranav Rajpurkar, a graduate student and co-lead author of the paper. “For example, two forms of the arrhythmia known as second-degree atrioventricular block look very similar, but one requires no treatment while the other requires immediate attention.”
In developing the algorithm the researchers collaborated with a wearable heartbeat monitor company, collecting a large dataset of unique ECG samples, which they then used to train a deep neural network model. In seven months, the algorithm was able to diagnose 14 types of arrhythmias about as accurately as cardiologists and outperform them in most cases.
“There was always an element of suspense when we were running the model and waiting for the result to see if it was going to do better than the experts,” says Rajpurkar. “And we had these exciting moments over and over again as we pushed the model closer and closer to expert performance and then finally went beyond it.”
The immediate potential of their algorithm, say the researchers, is for use in wearable devices that could provide real-time alerts of potentially dangerous heartbeat irregularities in at-risk patients. Looking ahead, the researchers hope the algorithm will be able to bring expert-level diagnoses to people in parts of the world who have no access to a cardiologist.
Wearable medical devices: Blurring the lines
Non-contact ECG monitors driver’s health
‘Heartbeat car’ displays driver’s pulse on biometric bodywork
Seatback driver alertness technology demonstrated
ECG front end IC shrinks cardiac monitoring wearables