Electrocardiogram screening for aortic valve stenosis using artificial intelligence

Comprehensive Summary

The aim of this study was to develop an artificial intelligence using convoluted neural networks to identify patients with moderate to severe aortic stenosis (AS). This model can be very beneficial, as moderate AS requires a different treatment than severe AS in order to be the most beneficial to the patient. A group of 258,607 patients who had an ECG taken between 1989 and 2019 using data from the Mayo Clinic database. They were then split into a training set, validation set, and testing set. The results of this created model showed that it had a high performance compared to current methods of detection which is the detection of the systolic murmur. It was also seen that the model’s accuracy increased when including factors of age and sex. This method is dependent on the skill of the physician and is likely only detected if the patient has symptoms. However, the AI model had a large model of false positives. These people were at higher risk for developing in AS and had other medical issues such as hypertension. This potential for early detection may allow for earlier treatment and hopefully better outcomes.

Outcomes and Implications

Early detection of aortic stenosis is important so that aortic valve replacement procedure can be done in time before a patient has severe symptoms which can lead to death. The AI model developed in this study can be used to help with earlier detection and detection in general in order to generate the best treatment plan for a patient. However, more testing should be done to ensure the model’s accuracy in differing populations.