Algorithm spots abnormal heart rhythms with doctor-like accuracy

From Engadget - July 8, 2017

See, an ECG does not always show the presence of arrhythmia, so doctors sometimes order patients to wear a wearable ECG device to monitor their heartbeat for two weeks. The researchers took advantage of that practice to create their algorithm. They used the massive dataset provided by heartbeat monitor company iRhythm to train a deep neural network for seven months until it's capable of detecting 14 types of arrhythmia by looking at ECGs.

To ensure the results' accuracy, the team had six cardiologists look at the data and diagnose each individual. They found that the algorithm was able to compete with the cardiologists, and in some cases, it was even able to differentiate between two very similar types of arrhythmia, something that could mean the difference between life and death.

Pranav Rajpurkar, the paper's co-lead author, said:


Continue reading at Engadget »