AusculThing ACC

The AusculThing ACC is a mobile-based decision support software that utilizes an AI-powered neural network to analyze heart sounds recorded via an electronic stethoscope, assisting healthcare providers in the evaluation of adult and pediatric patients. Its primary function is to automatically differentiate between normal and pathological heart sounds, specifically by detecting the presence of murmurs, with clinical data supporting its comparable performance to predicate devices in terms of accuracy, sensitivity, and specificity.

The AusculThing ACC is a decision support software designed for healthcare providers to aid in evaluating heart sounds in adult and pediatric patients. It functions by collecting acoustic heart signals via a Thinklabs One electronic stethoscope connected to a mobile device, on which the accompanying app runs; this app guides users in obtaining recordings from various body locations and allows for the phonocardiogram (PCG) data to be analyzed, visualized, and played back. The system's analysis capabilities stem from an AI-based algorithm, specifically a neural network model trained using a proprietary, echocardiogram-validated database of high-quality heart sound recordings. This algorithm automatically analyzes the collected sounds to distinguish between normal and pathological conditions, categorizing heart sounds as 'abnormal' if any murmur is identified or 'normal' if no murmurs or only benign murmurs are found, and provides information on suspected murmurs for abnormal findings. Clinical testing on a cohort of 133 patients (including 84 pediatric and 49 adult subjects, with 84 having confirmed heart defects) demonstrated the device's performance with an accuracy of 92.5%, sensitivity of 90.5%, and specificity of 96.0% in detecting abnormal heart sounds, using echocardiogram results as the gold standard; these results indicate its utility as a decision support tool, though it is not intended as a standalone diagnostic method.

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 features or types of murmurs does the AI algorithm identify beyond a general 'abnormal' classification, and does it differentiate between systolic and diastolic murmurs?; How does the proprietary echocardiogram-validated database used for training the AI account for variability in patient demographics (age, sex, BMI), comorbidities, and stethoscope placement techniques by different users?; What were the specific acoustic criteria and clinical definitions used to classify 'benign murmurs' as 'normal' by the AusculThing ACC algorithm?; How does the system's performance (sensitivity, specificity, accuracy) specifically vary between adult and pediatric populations, particularly in distinguishing innocent pediatric murmurs from pathological ones, given the combined validation data?; What safeguards or real-time feedback mechanisms are integrated into the mobile app to ensure optimal heart sound recording quality by healthcare providers with varying levels of auscultation expertise?