Skip to content

Algorithm diagnoses heart arrhythmias with cardiologist-level accuracy

In the near future, drones and wearables may combine forces to save a person experiencing a heart attack. Last month, a letter published in Journal of the American Medical Association detailed drones able to delivering AEDs to rural areas of Sweden faster than ambulances. Meanwhile, computer scientists from Stanford have shown promise in fulfilling the other piece of this solution.

In research published yesterday on arXiv and detailed in the video below, researchers from the Stanford Machine Learning Group, led by Andrew Ng, PhD, an adjunct professor of computer science, described their artificially intelligent algorithm that is able to outperform cardiologists in arrhythmia detection. Said Awni Hannun, a graduate student and co-lead author of the paper, in a Stanford News article about the work:

One of the big deals about this work, in my opinion, is not just that we do 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.

Developed over a period of seven months, the researchers trained the algorithm to differentiate between 13 types of arrhythmia using data from a wearable heart monitor produced by the company iRhythm. For their final tests, they had a group of cardiologists annotate 300 electrocardiogram (ECG) clips and then used those clips to compare the performance of the algorithm against cardiologists working alone. In the end, the algorithm’s diagnoses more closely matched those of the consensus group than the diagnoses of five individual cardiologists. Pranav Rajpurkar, a graduate student and co-lead author of the paper, described how it felt to reach this level of performance:

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. 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.

A significant advantage of this algorithm is that it is able to examine large amounts of data produced by wearable ECGs – often worn for two weeks straight – second-by-second, instantaneously and continuously. The researchers imagine this algorithm could someday be integrated into a wearable heart monitor, which would empower patients with easily accessible cardiologist-level assessment of their heart rhythms. Linked to a technology like the AED-carrying drone, their algorithm could also provide increased access to urgent treatment by alerting emergency services when a wearer is experiencing a potentially life-threatening arrhythmia.

Previously: Algorithm helps doctors guard patients against a second stroke and Atrial fibrillation more common than previously reported, study finds
Thumbnail photo by stux

Popular posts

Category:
Careers
Microaggressions in medical training: Understanding, and addressing, the problem

As a third-year medical student, Luisa Valenzuela Riveros, MD, was eager to begin participating in hospital rounds. But, as she told the audience at a Diversity and Inclusion Forum held Friday at Lucile Packard Children’s Hospital Stanford, one of her early case presentations didn’t go at all as she had hoped.
Category:
Nutrition
Busting myths about milk

Stanford nutrition scientist Christopher Gardner discusses the many forms of milk and addresses the biggest misconceptions.