The aim of the project is to develop an artificial intelligence software capable of analysing heart sounds to provide early diagnosis of a variety heart diseases at an early stage. Since the invention of the stethoscope by Laennec in 1816, the basic design has not changed significantly. Our software could be coupled with existing electronic stethoscopes to create an 'intelligent' stethoscope that could be used by healthcare assistants or practice nurses to screen for sound producing heart diseases. It could also be used at home by patients who would otherwise go undiagnosed.

The study investigators at Cambridge University Engineering Department (CUED) have developed a proof-of-concept AI algorithm to detect heart murmurs. However, in order to accurately detect the specific pathology and severity underlying the murmur, more heart sound recordings (matched with the ground truth from the patient's echocardiogram) are required. Patients presenting to one of the partner hospitals requiring an echocardiogram as part of their routine care will be invited to consent to this study. Participation will entail recording of a patient's heart sounds using an electronic stethoscope as well as collection of routine clinical data and a routine clinical echocardiogram at a single routine out patient visit.

This project will develop an AI algorithm which can be imported into a stethoscope to make it capable of automatically diagnosing any valve disease present and its severity. This will help GPs produce more accurate diagnoses, reduce costs by having fewer unnecessary referrals for echocardiogram, and produce more accurate diagnoses in countries where echocardiograms are not readily available due to their cost. Using a small sample of data as well as some which has been labelled by clinician auscultation, the team has created an award-winning AI algorithm capable of accurate detection of heart murmurs. However, in order to improve the accuracy and capability of this system more heart sound recordings from a range of diseases (matched with echocardiogram diagnosis) are required. The key to the success of this study will be to produce an AI algorithm that is more accurate than different grades of doctors at detecting the specific abnormality and severity underlying a heart murmur. This methodology will also provide a comprehensive study on acoustic characteristics of different heart sounds. So far all the acoustic characteristics of heart sounds taught to medical students are based on subjective opinion. This study will be able to objectively analyse these acoustic characteristics.

Inclusion Criteria:

• Participant willing and able to give informed consent for participation in study
• Participant to undergo an echocardiogram as part of their routine assessment

Exclusion Criteria:

• Informed consent is not given
• New York Heart Association (NYHA) functional class = 4

Patients attending cardiology clinics at the investigation sites who will be having an echocardiogram