eMurmur Heart AI

eMurmur Heart AI is a software-based decision support tool that utilizes an advanced algorithm to analyze digital heart sound recordings and patient age to assist clinicians in identifying and differentiating various heart sounds, including murmurs. Its methodology relies on processing 20-second auscultation data to detect S1/S2 sounds and classify murmurs as physiological, pathological, or absent, providing interpretations intended to augment, not replace, comprehensive clinical assessment.

The eMurmur Heart AI software serves as a decision support system for healthcare providers, designed to aid in the evaluation of recorded patient heart sounds. Its core function is the automated analysis of these sounds to identify S1 and S2 components, distinguish between physiological and pathological heart murmurs, and recognize the absence of a murmur. The system requires a 20-second digital recording of heart sounds, typically captured with a supported digital stethoscope such as the Littmann 3200, along with the patient's age. The analytical results are then presented on a front-end client, like the eMurmur app or web portal, for clinical review, emphasizing that the AI's interpretations should be used in conjunction with other patient information and professional medical judgment, not as a standalone diagnostic tool. The central technology of eMurmur Heart AI is its heart sound analysis algorithm, which is the regulated medical device component; the accompanying platform components (apps, web portal, backend for data handling) are considered non-medical device software. This algorithm is identical to the one used in a previous iteration, eMurmur ID, and therefore, eMurmur Heart AI shares the same established performance metrics for murmur detection: a sensitivity of 85.0% (Confidence Interval: 72.9% - 92.5%) and a specificity of 86.7% (Confidence Interval: 74.9% - 93.7%). The system processes auscultation data transmitted from the front-end client to the eMurmur backend, where the AI performs the analysis, and then returns the findings to the user to support their clinical decision-making process.

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 are the specific algorithmic features that enable the differentiation between physiological and pathological murmurs with the reported sensitivity of 85.0% and specificity of 86.7%?; How does the eMurmur Heart AI algorithm specifically utilize the patient's age, in conjunction with the 20-second heart sound recording, to improve analysis accuracy?; What validation was performed to ensure that the separation of the front-end/backend (non-medical device) from the AI algorithm (medical device) does not introduce risks to data integrity or analysis quality?; How does the system account for variations in recording quality stemming from different supported digital stethoscopes, user techniques, or ambient noise?; Can the performance metrics (sensitivity and specificity) be broken down for the individual identification of S1, S2, physiological murmurs, pathological murmurs, and the absence of murmurs, rather than just overall murmur detection?