Search Results

The eMurmur Heart AI software is a decision support system for the automated evaluation of recorded patient heart sounds. It identifies the first (S1) and second (S2) heart sounds, the presence of abnormal heart murmurs, and the absence of abnormal heart murmurs, with the latter including cases of no murmurs and innocent murmurs. eMurmur Heart AI also determines the average loudness of individual heart sound components, including S1, systole, S2, and diastole, as well as their respective loudness ratios.

eMurmur Heart AI is indicated for use in settings where auscultation can be performed properly by a healthcare provider. It is not intended as a sole means of diagnosis. The heart sound interpretations offered by eMurmur Heart AI are only significant when considered in conjunction with healthcare provider over-read and including all other relevant patient data. eMurmur Heart AI is intended for use on pediatric and adult patients.

To analyze heart sounds using eMurmur Heart AI, a digital recording of a patient's heart sounds is required. Recordings are made using any supported digital stethoscope, connected to a front-end client. The recorded auscultation data are transmitted from the front-end client to the eMurmur backend, which hosts eMurmur Heart AI. After analysis by eMurmur Heart AI, the results are returned to the front-end client where they are displayed to the user. The user can utilize the eMurmur Heart AI results to support their decision-making process regarding the potential presence of an abnormal heart murmur, and to assist them when investigating potential changes among longitudinally recorded heart sounds.

eMurmur is a multiple function software platform which includes the eMurmur apps and eMurmur web portal. The platform is used to stream, record, display, replay, and store auscultation data, recorded by means of supported digital stethoscopes, i.e., functions that are not subject to regulatory oversight.

The eMurmur multiple function software platform also has functions subject to FDA premarket review, i.e., eMurmur Heart AI. For this application, FDA assessed those functions only to the extent that they could adversely impact the safety and effectiveness of the functions subject to FDA premarket review.

Here is a summary of the acceptance criteria and study details for the eMurmur Heart AI (2.2) device, based on the provided FDA 510(k) clearance letter:

Acceptance Criteria and Device Performance

The study's goal was to demonstrate non-inferiority of eMurmur Heart AI (2.2) compared to its predicate (eMurmur Heart AI (1.0)) for sensitivity, specificity, and accuracy in identifying abnormal heart murmurs. Non-inferiority was established using confidence intervals for paired data with continuity correction for the true difference in sensitivity or specificity. The results shown above indicate that eMurmur Heart AI (2.2) met the non-inferiority criteria relative to the predicate device.

Study Details

1. Sample size used for the test set and the data provenance:

   • The exact sample size for the test set is not explicitly stated as a number of recordings. It is described as the "dataset used to validate and obtain clearance for the predicate device augmented with new, previously unseen data."
   • Data provenance: Not explicitly stated (e.g., country of origin). The data included heart sound recordings from both pediatric and adult subjects.

2. Number of experts used to establish the ground truth for the test set and the qualifications of those experts:

   • Number of experts: Three
   • Qualifications of experts: Board-certified cardiologists.

3. Adjudication method for the test set:

   • Method: Majority consensus. The ground truth classification ("abnormal murmur present" or "healthy") was determined by three board-certified cardiologists based on majority consensus.

4. If a multi reader multi case (MRMC) comparative effectiveness study was done, If so, what was the effect size of how much human readers improve with AI vs without AI assistance:

   • No MRMC comparative effectiveness study involving human readers and AI assistance was mentioned in the provided document for the primary validation of eMurmur Heart AI (2.2). The comparison was strictly between the eMurmur Heart AI (2.2) algorithm and the predicate algorithm (2.1).

5. If a standalone (i.e. algorithm only without human-in-the loop performance) was done:

   • Yes, a standalone (algorithm only) performance study was conducted. The "Performance Data" section directly reports the sensitivity, specificity, and accuracy of "eMurmur Heart AI (2.2)" and "eMurmur Heart AI (1.0)" (predicate) based on their automated analysis against the established ground truth.

6. The type of ground truth used:

   • Ground truth type: Expert consensus. Specifically, the "ground truth classification was determined by three board-certified cardiologists based on majority consensus."

7. The sample size for the training set:

   • The sample size for the training set is not explicitly stated in the provided document. The document mentions that the test set "was not used during training of eMurmur Heart AI (2.2)," implying a separate training set was used.

8. How the ground truth for the training set was established:

   • The method for establishing ground truth for the training set is not explicitly detailed in the provided document. It is only stated that the test set was not used for training. Given the approach for the test set, it is plausible that a similar expert consensus method was used, but this is not confirmed.

Ask a specific question about this device