Seeing the unseen: Machine learning in cardiac electrophysiology

This is a hybrid meeting. Please find the Teams link in the abstract.

Dr Rasheda Chowdhury is an advanced research fellow and group lead at Imperial College London, within the National Heart and Lung Institute. She is a member of the member of the ElectroCardioMaths Programme within the Imperial Centre for Cardiac Engineering.

Cardiac arrhythmias are a group of conditions in which the electrical activation and consequent contraction of the heart is irregular, or faster or slower than normal, preventing the heart from carrying out its role of sustaining circulation. Abnormal electrical conduction through the heart can lead to arrhythmias. The contact electrogram is recorded from electrodes placed inside the heart and can be used clinically to locate areas of abnormalities, as they depict the local electrical activity of the myocardium. They can be used to guide therapy for arrhythmias, namely ablation procedures, where the area of the heart which is triggering or sustaining the arrythmia (the substrate) is destroyed to prevent its abnormal activity. However, the detailed content of the electrogram and associated pathophysiological changes are not well understood. For this reason, the outcomes of ablation strategies for complex remains low. This leads to a health burden for patients and a financial burden for the NHS.

Signal processing, supervised machine learning and neural networks may be able to guide ablation procedures if determined to be effective at localising the pathological substrate. These predictive models can be developed, then tested, on in silico and laboratory models, where the ground truth is more easily identifiable, to determine their efficacy. This talk will outline Dr Chowdhury’s work to date in developing machine learning models to predict the underlying substrate from the contact electrogram.

Teams link: