A self-supervised machine learning model for identifying abnormal heart rhythms

Project Reference :


Institution :

National University of Singapore (NUS), Singapore General Hospital (SGH)

Principal Investigator :

Prof Feng Mengling

Technology Readiness :

3 (Experimental proof of concept)

Technology Categories :

AI- Machine Learning

Background/Problem Statement

Cardiovascular disease (CVD) is a leading global cause of death. 12-lead electrocardiography (ECG) is pivotal for heart health checks as it aids in the detection of irregular heart rhythms (cardiac arrhythmias) that could indicate serious heart issues leading to CVD. However, for training machine learning models to automatically detect cardiac arrhythmias, a ton of precisely labelled data is required which is often lacking despite the vast amounts of collected ECG data. Labelling this data is time-consuming and costly, posing a significant challenge for real-world clinical use.


Developed in close collaboration with SGH, the intra-inter self-supervised machine learning model (“ISL”) is an end-to-end solution designed to improve the identification of abnormal heart rhythms from unlabeled multivariate cardiac signals. It incorporates two distinct self-supervision procedures: intra-subject and inter-subject self-supervision. Both procedures incorporate medical domain knowledge to enhance the diagnostic performance of cardiac arrhythmias, offering significant value in real-world applications.

Link to publication: intra-inter self-supervised machine learning model


Extensive experiments, conducted on three public datasets to ensure reproducibility, reveal that the ISL surpasses current state-of-the art methods in the downstream tasks across various scenarios.

With approximately a 10% improvement over supervised training when only 1% labeled data is available, the technology demonstrates its ability to identify cardiac arrhythmias effectively, showcasing adaptability in situations with limited information.

Potential Application(s)

Healthcare facilities conducting heart health assessments can utilize this solution to identify cardiac arrhythmias, effectively cutting down on costs and diagnostic time to provide timely and appropriate care to patients.

We welcome interest from the industry for collaboration/ co-development / customisation of the technology into a new product or service. If you have any enquiries or are keen to collaborate, please contact us.