Tyto Insights for Rhonchi Detection

Tyto Insights for Rhonchi Detection is an AI-enabled prescription-use software system that analyzes acoustic lung sound recordings from the FDA-cleared Tyto Stethoscope to detect abnormal sounds suggestive of rhonchi in patients aged 2 years and older. The device provides a positive, negative, or unable-to-analyze result and requires a licensed healthcare professional to interpret the findings in conjunction with other clinical data.

The device is a web-based AI algorithm with three subsystems: an Application Server, an Algorithm Server that performs the acoustic analysis, and a Web Server providing graphical results. It processes recordings in segments of 6–12 seconds and delivers binary classification outputs (rhonchi suspected or not). Performance shows sensitivity of 0.60 (95% CI: 0.50–0.69), specificity of 0.99 (95% CI: 0.97–1.00), and AUC of 0.96 (95% CI: 0.92–0.98) on a retrospective validation dataset of 400 recordings.

ANSI AAMI ISO 14971:2019 (risk management), ANSI AAMI IEC 62304:2006/A1:2016 (software lifecycle), ISO 15223-1 (medical device symbols), and ANSI AAMI IEC 62366-1:2015+AMD1:2020 (usability engineering). The submission also references FDA guidance on cybersecurity and software verification/validation.

The device is substantially equivalent because it shares the same intended use (AI-enabled detection of specific abnormal lung sounds in patients 2+ years old), the same patient population, the same input signal type (Tyto Stethoscope recordings), the same user type (HCP and lay users), the same regulatory classification (Class II, product code PHZ, 21 CFR 868.1900), and the same output format (indication of abnormal sound detected or not). Although it targets rhonchi rather than crackles (as the predicate does), this difference does not raise different safety or effectiveness questions, as both are specific abnormal breath sound detections requiring HCP interpretation. Clinical validation demonstrated non-inferiority (AUC 0.96 vs. clinical readers 0.79) with consistent performance across age, sex, and user subgroups, establishing equivalent diagnostic capability.