Our research aims to leverage machine learning methods to assist in the diagnosis of Atrial Fibrillation. Atrial Fibrillation is one of the most common arrhythmias and is associated with a higher risk of stroke. It occurs in episodes and when active, can be identified by its abnormal rhythm on an electrocardiogram (ECG). However, often there is no active episode, when the patient is at the doctor’s office and diagnosis requires long-term monitoring with portable ECG machines at home. Using machine learning to identify the disease, even when the current rhythm of the heart appears as healthy on the ECG, could lead to reduced time to diagnosis and therefore, ease the burden and risk of stroke in patients.
During the three year duration of the project we plan to research different machine learning methods and find the best algorithms for processing ECG data. Furthermore we plan to research different means to explain the results from the machine learning model to the clinical personal and patients making the results reliable and trustworthy. In order to achieve this, we plan to make use of the inherent geometrical structure of the signals from the different ECG electrodes, as well as expert knowledge from doctors.
