Challenging the Limitations of Atrial Fibrillation Detection in the Presence of Other Cardiac Arrythmias

Area of Science:

• Cardiology
• Biomedical Engineering
• Signal Processing

Background:

• Atrial fibrillation (AF) diagnosis is crucial for timely treatment, yet current wearable devices often fail to differentiate AF from other arrhythmias.
• Existing rhythm detection methods primarily distinguish between normal and abnormal heart rhythms, lacking specificity for conditions like AF.

Purpose of the Study:

• To evaluate the performance of a traditional inter-beat interval (IBI) based classifier in distinguishing atrial fibrillation (AF) from sinus rhythm (SR) when other cardiac arrhythmias are present.
• To quantify the impact of non-AF arrhythmias on the accuracy of an IBI-based AF detection algorithm.

Main Methods:

• Utilized data from 37 patients with optical heart rate monitors during catheter ablation procedures.
• Assessed a traditional classifier's performance on pure AF vs. SR data, then progressively introduced non-AF arrhythmias into the classification windows.
• Analyzed classification metrics including accuracy, sensitivity, and specificity.

Main Results:

• The classifier achieved high performance (accuracy 0.979, sensitivity 1.000, specificity 0.966) for pure AF vs. SR detection.
• Performance decreased significantly when non-AF arrhythmias constituted up to 50% of the data (accuracy 0.886, sensitivity 0.998, specificity 0.853).
• Sinus tachycardia caused false positives, but other arrhythmias like extrasystoles and atrial flutter did not impair classification.

Conclusions:

• Traditional IBI-based classifiers are insufficient for reliably distinguishing AF from a variety of other cardiac arrhythmias.
• Future research requires diverse arrhythmia datasets and novel classification features to improve AF detection accuracy in real-world wearable devices.