FDA-Cleared, AI-Assisted ECG Analyzer Algorithm Reduces False Positives in ICMs

On July 28, the US Food and Drug Administration (FDA) granted 510(k) clearance to the ILR ECG Analyzer, which serves as a second layer for residual alerts generated by implantable cardiac monitors (ICMs).1

ICMs are well-known for their tendency to produce false positive alerts, detecting arrhythmias that are not present or clinically significant. Even with the implementation of AI algorithms to filter these alerts from genuine arrhythmias, the substantial burden posed by false positives still lingers. Past research has shown that many devices register benign rhythms or certain signal artifacts, such as electrical noise or premature ventricular contractions, which are frequently labeled as significant events.2

To combat this, Implicity’s ILR ECG Analyzer was developed to act as a backup filter for alerts that bypass initial manufacturer filters. The algorithm can analyze data from all major device manufacturers. Recent research has shown that the algorithm reduced false positives by ≤76% while maintaining high sensitivity for detecting clinically meaningful events.1

“Essentially, you keep looking at something – if it’s real, that’s great, but you get so much information that’s not real that it creates a major workflow challenge,” Manish Wadhwa, MD, a cardiologist and electrophysiologist with San Diego Arrhythmia Associates and chief executive officer of SummaCor, told HCPLive in an exclusive interview. “That’s why something like this, this AI-based analyzer, is so important: it helps to mitigate the workflow challenges.”

In a recent study published in JACC, investigators conducted a multicenter, multivendor database analysis of ICMs to compare the accuracy of arrhythmia detection between 4 commonly used ICMs – the LINQ II from Medtronic (MDT), the LUX-Dx from Boston Scientific (BSX), the Biomonitor III from Biotronik (BIO), and the Confirm Rx from Abbott (ABT). A total of 437,351 electrocardiogram-verified episodes were collected from 6756 patients across 25 centers – investigators selected a random sample of 1140 patients based on a prospectively determined power analysis.3

The study cohort generated 25,826 total alerts, with a median of 5 alerts (interquartile range [IQR], 2-14) per patient. ICM implant indications included cryptogenic stroke (21%), suspected atrial fibrillation (11.3%), atrial fibrillation management (23.6%), syncope (23.2%), palpitations (14.9%), ventricular tachycardia (2.1%), and unknown (3.8%).3

Atrial tachycardia and atrial fibrillation were the most frequent alerts (BIO, 70.8%; MDT, 68%; ABT, 55.7%; BSX, 35.7%). After adjudication, BSX displayed the highest positive predictive values, followed by ABT and MDT, while BIO had the lowest (BSX, 0.73; MDT, 0.56; ABT, 0.51; BIO, 0.23). False positives for atrial tachycardia and atrial fibrillation were attributed to ectopy (75.8%), combined ectopy with oversensing or noise (5.3%), isolated oversensing (5.2%), sinus arrhythmias (3.9%), indeterminate reason (3.7%), noise (3.6%), and bradycardia (2.5%).3

BSX demonstrated the highest positive prediction values (0.96), while MDT had the lowest (0.36) – this was primarily attributed to undersensing. ABT had the lowest pause detection accuracy (positive prediction values: 0.01; 79.8% due to undersensing) and BSX had the highest (positive prediction values: 0.7). Additionally, tachycardia alerts had moderate positive prediction values for BIO (0.62) and BSX (0.61), with substantially lower values for MDT (0.29) and ABT (0.27).3

“There is one other benefit to this as well, and that is standardization. If you took a single strip and you asked 20 different practitioners to interpret that strip, sadly, you’re probably going to get about 10 to 15 different answers,” Wadhwa said. “To that end, if there is a ground truth, what you actually want is for the standardized output to be the truth. You can keep asking people and you’ll keep coming back with different answers, but the AI is very sure, and if we train the models properly, it will always come back the same way.”

Editors’ Note: Wadhwa reports no relevant disclosures.

References

1. Implicity. Implicity’s Next-Generation ILR ECG Analyzer Earns FDA Clearance. GlobeNewswire. July 28, 2026. Accessed August 21, 2026. https://www.globenewswire.com/news-release/2026/07/28/3334149/0/en/implicity-s-next-generation-ilr-ecg-analyzer-earns-fda-clearance.html

2. Implicity. EHRA 2026 Studies Reveal Why False Positives Persist in AI-Equipped Implantable Cardiac Monitors. April 14, 2026. Accessed August 21, 2026. https://implicity.com/ehra-2026-studies-reveal-why-false-positives-persist-in-ai-equipped-implantable-cardiac-monitors/

3. Kamsani SH, Middeldorp ME, Evans S, et al. Accuracy of implantable loop recorders. JACC: Clinical Electrophysiology. 2026;12(5):1113-1127. doi:10.1016/j.jacep.2025.12.039