A Novel Atrial Fibrillation Prediction Algorithm Applicable to Recordings from Portable Devices

Area of Science:

• Cardiology
• Biomedical Engineering
• Data Science

Background:

• Atrial Fibrillation (AFib) is a significant risk factor for severe cardiovascular conditions.
• Prevalence of AFib in the U.S. is substantial (2.7-6.1 million) and projected to rise with an aging population.
• Accurate and timely detection of AFib events is crucial for patient management.

Purpose of the Study:

• To develop and validate an automated method for labeling symptomatic Atrial Fibrillation events.
• To utilize portable ECG recordings for AFib detection.
• To improve the efficiency of identifying AFib episodes.

Main Methods:

• A novel heart rate-duration criteria region was defined for AFib event identification.
• A Markov Chain algorithm was employed to classify 2-minute prediction intervals preceding symptomatic AFib.
• The algorithm was tested using data from portable ECG monitors.

Main Results:

• The proposed method achieved an overall accuracy of 82% in classifying AFib events.
• The Area Under the Curve (AUC) for the classification was 0.91, indicating strong predictive performance.
• The approach demonstrated effectiveness in automatically labeling symptomatic AFib episodes.

Conclusions:

• The developed automated method shows promise for accurate detection of symptomatic Atrial Fibrillation.
• This technique, using ECG data and a Markov Chain algorithm, can aid in the early identification of AFib.
• Further validation is warranted, but the preliminary results suggest a valuable tool for managing AFib patients.