Early screening of your heart valves

We are combining world-leading AI and acoustics to enable widespread early screening of valvular heart disease, preventing costly late detection of heart failure.



1 in 9

over 65s have significant valvular heart disease. This is nearly 2 million people in the UK and will grow rapidly due to the ageing population.

50%

of patients are undiagnosed and will progress to late stage where the prognosis is worse than advanced stage cancer.

£345m

additional cost to the NHS per year because of treating patients at a later, symptomatic stage.

The current detection pathway for VHD is inadequate

1. Symptoms

Patients mistake the classic symptoms of valvular heart disease (breathlessness, fatigue) for other conditions or just the natural process of ageing. As a result, they wait until a late stage before contacting the healthcare system.

2. GP stethoscope exam

Patients with VHD symptoms are infrequently examined with a stethoscope. Even when a stethoscope is used, accurate diagnosis of VHD from sounds requires many years of training and experience. GPs miss more than half of clinically significant cases.

3. Echocardiography

A heart ultrasound is the gold-standard test, but requires a highly skilled operator to both acquire and interpret the images. Hospitals have limited staff and a majority of GP referrals are unnecessary.

Our technology: AI-enabled screening of valve disease
in humans and animals

z

Novel acoustic hardware

A new medical device and advanced signal processing to make recording heart sounds from the body simple and straightforward [3].

i

Leading clinical data

High-quality clinical datasets collected through bespoke studies in the NHS and veterinary centres in the UK.

6500+ recordings from humans [4]

2500+ recordings from dogs [2]

T

AI-driven diagnoses

Award-winning [1] machine learning techniques to automatically analyse sound recordings and detect and grade the presence of valvular heart disease.

93% sensitive to human
heart murmurs [1]

88% sensitive to canine
heart murmurs [2]

Backed by published research:

[1] McDonald A, Gales MJF, Agarwal A. A recurrent neural network and parallel hidden Markov model algorithm to segment and detect heart murmurs in phonocardiograms. PLOS Digit Health. 2024 Nov 25;3(11):e0000436. doi: 10.1371/journal.pdig.0000436.

[2] McDonald A, Novo Matos J, Silva J, et al. A machine-learning algorithm to grade heart murmurs and stage preclinical myxomatous mitral valve disease in dogs. J Vet Intern Med. 2024 Nov-Dec;38(6):2994-3004. doi: 10.1111/jvim.17224.

[3] McDonald A, Nussbaumer M, Rathnayake N, Steeds RP, Agarwal A. A flexible multi-sensor device enabling handheld sensing of heart sounds by untrained users. Under review. Preprint available at medRxiv. doi:10.1101/2024.10.09.24315183.

About us

We are based in the University of Cambridge Acoustics Lab. We are a team of engineers, machine learning experts and experienced clinicians who are all passionate about transforming patient lives by improving early detection of disease across the world.

We collaborate with leading health systems, veterinary centres, and companies to develop impactful solutions.