Novel artificial intelligence algorithms reduce electrogram burden and arrhythmia alerts imposed by Assert-IQ insertable cardiac monitors
R Gopinathannair, D Yoo, M Katcher, P Katrapati, F Qu, F Dawoud, D Monje, K Davis, A Goil, D LakkireddyAbstract
Background
Insertable cardiac monitors (ICMs) are essential for ambulatory arrhythmia diagnosis. Two novel artificial intelligence (AI) based algorithms have been developed recently for automatic classification of ICM documented atrial fibrillation (AF) and pause electrogram (EGM). The algorithms were designed to reduce false episodes and alerts while maintaining high sensitivity of detecting true AF and pause event.
Purpose
To evaluate the impact of these AI algorithms on EGM and alert burden in real world patients implanted with Abbott Assert-IQ™ ICMs.
Methods
Retrospective analyses of Assert-IQ devices that have not been utilized in the AI algorithm development and with >90 days of remote transmission history were performed. Data from remote transmissions within 90 days since the first transmission on Merlin.net™ patient care network were deidentified and extracted for the analyses. The AI algorithms and associated alert modification logic were applied to each transmission, and the clinical impact on EGM and alert burden reduction was evaluated. According to the alert modification logic, AF or pause alert in a device transmission was removed if all AF or pause EGM episodes in the transmission were classified as false positive by the AI algorithms.
Results
The dataset included 4,042 devices in total. Patient reason for monitoring included syncope (30%), AF-related (37%), cryptogenic stroke (22%) and other reasons (12%). Prior to AI classification, the frequency of device transmissions containing AF or pause alert was 3.5 (2.9 for AF, 0.7 for pause) per patient month. The AI algorithms reduced AF EGM data burden by 41% and pause EGM data burden by 83%, and reduced the frequency of transmission containing AF or pause alert to 2.1 (2.0 for AF, 0.2 for pause) per patient month.
Conclusion
The novel AI algorithms significantly reduce AF and pause EGM and alert burden imposed by Assert-IQ devices. Implementing these models on the remote monitoring server may reduce clinical cost and time required for arrhythmia diagnosis and management.