Ex-post correction of pacemaker mode switch episodes in undersensed atrial fibrillation

Area of Science:

• Biomedical Engineering
• Cardiology
• Data Science

Background:

• Cardiac implantable electronic devices (CIEDs) collect data on atrial fibrillation (AF) episodes.
• CIEDs exhibit inaccuracies in detecting short AF episodes, impacting diagnostic reliability.
• Mode switching events correlate with long AF episodes but are error-prone for short ones.

Purpose of the Study:

• To develop and validate a novel algorithm for accurate detection of short AF episodes from CIED data.
• To improve the diagnostic evaluation of paroxysmal arrhythmias by enhancing AF episode detection.
• To refine data analysis from CIEDs without altering device sensitivity parameters.

Main Methods:

• Expectation-maximization algorithms were employed to estimate parameters from CIED-recorded AF episodes.
• The method addresses missing episode durations and merges short episodes into longer events.
• Post-processing focused on identifying false negatives, ensuring safer arrhythmia diagnostics.

Main Results:

• The algorithm modified the final data histogram in 40 out of 76 patients studied.
• An average of 2.79% of episodes shorter than 1 minute were removed.
• 1% of previously unaccounted episodes were identified as longer than 30 minutes, with 16% exceeding 24 hours.

Conclusions:

• The developed method is stable and reliably preserves the detection of long arrhythmia episodes.
• The algorithm demonstrates high similarity to human expert analysis in identifying new long episodes.
• This approach enhances the safety and accuracy of diagnosing paroxysmal arrhythmias.