Powering Hearts
Beat to Beat

Q: AI means so many different things in 2026.  Can you tell us more about the AI models powering the Assert-IQ ICM System?

The Assert-IQ ICM is a paperclip-sized device that gets implanted under the skin on a patient’s chest, where it monitors the heart’s electrical signals for several years. While device-based arrhythmia detection algorithms have advanced significantly over the past several years, there were still too many false positives being sent to healthcare providers alongside true episodes.  We want our providers spending time where it matters, engaging with patients and responding to those that need care, instead of managing devices and data burden. As a result, we innovated with powerful, cloud-based AI algorithms to classify these episodes, improve accuracy, and limit the data deluge.  This upgraded Assert-IQ ICM platform improves clinical efficiency while maintaining the high arrhythmia sensitivity that is critical to providers.

For more specifics around the technology, our AI algorithms are neural networks (mathematical models) that leveraged supervised machine learning to “learn” from thousands of labeled electrograms (EGMs) graded by electrophysiologists.  By learning from this real-world, expert-verified “ground truth”, the models reliably recognize the same patterns and classify episodes in a way that aligns with how specialists read EGMs.

Q: Can you elaborate on why this matters for providers and device clinics?

With AI classification working before data reaches the clinic review queue, the Assert-IQ ICM system transforms a stream of device detections into a more curated, decision-ready view. This is about delivering fewer, more actionable notifications so providers can redirect time from sorting alerts to acting on insights, reducing data fatigue without compromising vigilance. The result is a clearer path from continuous monitoring to timely clinical action. 

Q: What would you tell more skeptical cardiologists or patients about Assert-IQ ICM’s new AI functionality?  Why should they trust it?

Let’s talk first about our regulatory process and algorithm training, and then we can review new clinical evidence confirming device and AI performance.

We built these AI algorithms with patient safety at the center, engaging with the FDA throughout development and locking the models at FDA submission for predictable, auditable performance. This is deterministic AI, not generative AI, so it delivers the same output for the same input and does not “hallucinate.”

The training set included thousands of real‑world EGMs from a diverse patient population across more than 1,000 clinics, with experts labeling whether AF or Pause episodes were occurring. We then validated performance on a large, independent, expert‑labeled cohort with no exclusion criteria to reflect real‑world variability. In model evaluation, AI cut false positives by 81% for AF by 91% for Pause, while maintaining high episode sensitivity.

Finally, we recently assessed Assert-IQ ICM AF detection and AI performance in a prospective, multicenter study.  It enrolled 151 patients to compare Assert-IQ ICM AF detection against Holter monitoring to assess system performance.  The study found that Assert-IQ ICM demonstrated 100% patient sensitivity, 99.4% episode-based sensitivity, and that the AI algorithm retained all true positive AF detections while eliminating false detections.

AI now elevates atrial fibrillation management by ensuring that more of the right data is analyzed and summarized for action. The system can transmit up to approximately 60 minutes of recordings to Merlin.net™ Patient Care Network (PCN) without affecting battery life, which means the AI sees and classifies more signals reducing the review needed by clinics. Electrophysiologists then receive periodic summaries, including AF burden trends, so they can assess trajectory at a glance and act when it matters. The net result is less time hunting for insights and more time applying them, with sensitivity for AF detection preserved as part of the platform’s design.

Looking ahead, we’re excited by how long‑term monitoring and AI‑driven classification come together in Assert‑IQ ICM to raise the bar for continuous rhythm assessment. The platform’s six‑year battery longevity* supports extended monitoring windows, while cloud-based AI ensures that the data you see is more meaningful with zero impact on device battery life. Building on this foundation, we see an opportunity to extend learnings into long‑term disease monitoring and to explore signals related to progressive conditions to support earlier interventions.

Q: Can we expect AI to impact other areas of cardiac care at Abbott in the future?

Yes. The same disciplined, evidence‑driven approach used for Assert‑IQ ICM opens the door to broader impact across cardiac care. We see potential to enhance remote workflows with smarter prioritization, to offer guidance that helps teams converge on device settings and to explore multimodal signals that may surface clinically relevant trends earlier, pending rigorous validation and regulatory review, and we’re excited to maintain this pace of innovation.

*DM5500 model.

™ Indicates a trademark of the Abbott group of companies. 

‡ Indicates a third-party trademark, which is property of its respective owner. 
Bluetooth and Bluetooth logo are registered trademarks of Bluetooth SIG, Inc.