Powerful Deep Learning Algorithm

Expert-level accuracy
Accelerated patient care

Our goal

Helping healthcare professionals achieve a definitive diagnosis, and accelerate your patients treatment pathway

Combining deep learning and classic machine learning capabilities with our expert cardiac technician review, Zio by iRhythm gives physicians the assurance of expert-level accuracy in arrhythmia detection while bringing patient care to the forefront.

10 Arrhythmia classifications with expert accuracy

Streamlined workflows for optimal efficiency

More time spent with patients, not with patient data

Our algorithm

Expedition of large ECG data sets increase efficiency at scale

AI technology can expedite comprehension of large ECG data sets, boosting clinical efficiency and letting physicians focus on patient care rather than having to re-look at questionable strips.

Quality data, robust analysis

99% mean analysable time

The Zio Service provides uninterrupted data and high diagnostic yield to diagnose arrhythmias accurately first time. Turakhia, M., et al. Diagnostic Utility of a Novel Leadless Arrhythmia Monitoring Device. The American Journal of Cardiology, 2013

deep learning

Our collaboration with the Stanford Machine Learning Group shows that the Zio service can provide the reliability of an expert cardiologist. This could only only be achieved because of the large amounts of diverse data generated by the Zio Service.

Certified Cardiac Technician Review

We employ certified cardiac technicians
specialised in advanced arrhythmia interpretation. Our experts review all reports prior to posting, giving confidence and assurance with the Zio Service.

In collaboration with the Stanford Machine Learning Group, over 91,000 Zio records were processed through a deep neural network (DNN) to see if the algorithm could learn to identify various arrhythmia classes.

The quality and quantity of Zio data enabled the DNN to accurately detect and identify 10 arrhythmia classes while separating sinus rhythm and artifact for a total of 12 output classes. These results were compared to those of board-certified experts, confirming equivalent performance.

To learn more about the collaboration, visit the Stanford Machine Learning Group online