IRNF App

The Irregular Rhythm Notification Feature 2.0 (IRNF 2.0) is a mobile application designed to opportunistically screen for signs of atrial fibrillation in adult users without a prior AFib diagnosis, utilizing photoplethysmography data captured by an Apple Watch. Its methodology relies on a proprietary convolutional neural network to analyze pulse rate patterns for irregularities suggestive of AFib, subsequently providing notifications to encourage users to seek further medical evaluation.

The Irregular Rhythm Notification Feature 2.0 (IRNF 2.0) by Apple Inc. is a software-only mobile medical application designed for over-the-counter use with compatible Apple Watch and iPhone devices. Its purpose is to analyze pulse rate data, collected by the Apple Watch's photoplethysmography (PPG) sensor when the user is still, to identify episodes of irregular heart rhythms suggestive of atrial fibrillation (AFib) and provide a notification. The system comprises an Apple Watch app, which contains a new machine learning-based rhythm classification algorithm (a convolutional neural network), a confirmation cycle algorithm, and notification generation capabilities, while the paired iPhone app provides onboarding, educational materials, and a history of AFib notifications. This feature functions as an opportunistic background screening tool, intended for individuals aged 22 years or older who have not been previously diagnosed with AFib, and aims to supplement AFib screening decisions rather than replace traditional methods of diagnosis or treatment. The underlying science for IRNF 2.0 involves a convolutional neural network that was trained extensively using over 3 million pulse rate recordings from more than 2500 demographically diverse subjects, encompassing various rhythms including atrial fibrillation, normal sinus rhythm, sinus arrhythmia, and other ectopic beats. Clinical performance validation demonstrated that IRNF 2.0 achieved person-level sensitivity of 88.6% and specificity of 99.3% when compared to a predicate device on a common sensor dataset. In a separate clinical study with 573 participants (ages 22 and older) who wore an Apple Watch and a reference electrocardiogram (ECG) patch, the IRNF 2.0 showed, for the primary analysis cohort (n=432), a sensitivity of 88.6% in detecting AFib (124 out of 140 subjects with ECG-confirmed AFib received a concordant notification) and a specificity of 99.3% (290 out of 292 subjects without AFib on ECG did not receive a notification). Another study involving 226 participants found that of 57 who received AFib notifications, 78.9% showed concordant AFib on the ECG patch, and 98.2% showed AFib or other clinically relevant arrhythmias.

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. How does the performance of the convolutional neural network in IRNF 2.0 vary across different demographic subgroups, particularly different skin tones, given its reliance on PPG sensor data?; What is the minimum duration and quality of 'stillness' required for the PPG sensor to reliably collect data for the IRNF 2.0 algorithm to perform an analysis?; Considering the 78.9% concordance with ECG-confirmed AFib in the 226-participant study, what were the characteristics of the discordant cases (i.e., instances where IRNF 2.0 notified but ECG did not show AFib, or other arrhythmias identified)?; How does the 'confirmation cycle algorithm' function, and what are its specific criteria for confirming an irregular rhythm before issuing a notification to the user?; What specific 'other clinically relevant arrhythmias' were detected in the 98.2% of notified participants in the 226-person study, and how does the device differentiate or flag these if its primary indication is for rhythms suggestive of AFib?