2 min read

A group of researchers from Stanford University and University of California with iRhythm Technologies Inc. and Veterans Affairs Palo Alto Health Care System have build a model that can help in the diagnosis of irregular heart rhythms, also called as arrhythmias.

On Monday, the researchers shared their findings in a paper published on Springer Nature: Cardiologist-level arrhythmia detection and classification in ambulatory electrocardiograms using a deep neural network.

Detecting arrhythmias is a pretty easy task for an expert technician or a cardiologist but is known to be quite challenging for computers. With the help of widely available ECG data and deep learning, this study aimed to improve the accuracy and scalability of automated ECG analysis.

For this study, the researchers built a 34-layer deep neural network (DNN) and trained it to detect arrhythmia in arbitrary length ECG time series. The model was trained on 91,232 single-lead ECGs from 53,549 patients who used a single-lead ambulatory ECG monitoring device. The network learned to classify noise and the sinus rhythm. Additionally, it also learned to classify and segment twelve arrhythmia types present in the time series.


For testing the model, the researchers used an independent test dataset annotated by a consensus committee of board-certified practicing cardiologists. The test dataset used in this study is publicly available at iRhythm Technologies’s GitHub repository.

The model did pretty well by achieving an average area under the receiver operating characteristic curve (ROC) of 0.97. Another measure of accuracy was F1, which is a harmonic mean of the positive predictive value and sensitivity. F1 score of the DNN (0.837) exceeded that of average cardiologists (0.780).

Read Next

Researchers introduce a CNN-based system for identifying radioresistant cancer-causing cells

Stanford researchers introduce DeepSolar, a deep learning framework that mapped every solar panel in the US

Our healthcare data is not private anymore: Study reveals that machine learning can be used to re-identify individuals from physical activity data