LINQ II Insertable Cardiac Monitor, Zelda AI ECG Classification System

The LINQ II Insertable Cardiac Monitor system, now enhanced with the Zelda AI ECG Classification System, aims to provide continuous, long-term subcutaneous ECG monitoring for arrhythmia detection in adult patients. The core innovation lies in the Zelda AI's use of deep-learning neural networks to improve the accuracy of detecting Atrial Fibrillation and Pause episodes by significantly reducing false alerts while maintaining true event capture, with future algorithm updates managed by a structured Predetermined Change Control Plan.

The LINQ II Insertable Cardiac Monitor (ICM) Model LNQ22 is a programmable, implantable device designed for continuous subcutaneous ECG monitoring in patients at risk of cardiac arrhythmias or experiencing transient symptoms suggestive of such conditions, like dizziness, palpitations, syncope, or chest pain. It automatically records cardiac events and allows patient-initiated recordings via a mobile app or a dedicated patient assistant. The system classifies arrhythmias into categories such as tachyarrhythmia, bradyarrhythmia, pause, atrial tachyarrhythmia, and atrial fibrillation. The core LINQ II ICM hardware, constructed with materials like sapphire and titanium for a hermetic enclosure and reliable sensing, remains technologically unchanged from its predicate device. The key enhancement detailed is the introduction of the Zelda AI ECG Classification System (Models ZA400, ZA410, ZA420), which integrates with the LINQ II ICM. The Zelda AI system employs deep-learning neural architectures, specifically residual convolutional neural networks and ensemble models, to improve the detection of Atrial Fibrillation (AF) and Pause episodes. The primary scientific purpose of the Zelda AI is to significantly reduce false alerts generated by the ICM for these specific arrhythmia events, while ensuring the retention of true alerts. Performance validation testing confirmed its effectiveness in meeting these criteria. Future algorithm improvements for the Zelda AI system are governed by a Predetermined Change Control Plan (PCCP), allowing for modifications such as threshold adjustments and algorithm re-training under strict protocols, with any changes locked before deployment and aimed at achieving superior, validated performance.

AI Regulatory Concerns

Powered by

Jev reads the full FDA letter and returns calibrated concern probabilities per regulatory category. These are AI estimates, not regulatory advice.

• Q1. What specific performance metrics (e.g., sensitivity, specificity, positive predictive value) were achieved by the Zelda AI ECG Classification System in reducing false alerts for AF and Pause episodes compared to the predicate LINQ II ICM without Zelda AI?; How does the Zelda AI algorithm differentiate between true cardiac events and noise or artifacts that commonly lead to false alerts in ICMs?; What are the defined thresholds and data characteristics used for re-training and pre-training the Zelda AI algorithm under the Predetermined Change Control Plan, and how is "superior performance" quantified for these updates?; Given the LINQ II ICM is not tested for pediatric use, are there plans to validate the Zelda AI system's performance in pediatric populations or specific subpopulations with unique arrhythmia profiles?; What is the computational overhead of the Zelda AI ECG Classification System, and does its operation impact the battery life or data storage capacity of the LINQ II ICM?