Data Labeling for Medical Audio

Audio clips from medical devices can be powerful indicators of a patient’s health and can be leveraged to train predictive algorithms to detect heart conditions, lung abnormalities, and more.

We have annotation and classification services tailored specifically to medical audio clips, with features such as range selection, spectrogram integration, and motion data integration. Our work includes identifying cough and wheeze, classifying heart murmurs, and detecting the number of abnormal breath sounds. 

Data Types

Audio
Audio
Audio

Case Studies | 

Data Labeling for Medical Audio

Eko

Eko

Using recordings from their digital stethoscopes, Eko Health integrates machine learning algorithms into their products to detect heart and lung abnormalities. Centaur has identified murmurs, coughs, wheezes, rhonchi, and more at a tremendous scale. In one task, Centaur gathered over 120,000 opinions in just two weeks - and the new data increased Eko’s AI model AUC from 0.87 to 0.92. Learn more about our work with Eko here.

Feebris

Feebris

Feebris is on a mission to empower a community workforce to deliver precise and effective medical care through AI. They have enlisted Centaur Labs to assist in their creation of an AI algorithm that can be used to detect abnormalities in respiration. Centaur has segmented ranges of clear respiration in thousands of audio clips, labeling up to 600 clips per day. 

Customer Testimonials

I was personally really impressed by Centaur's approach to annotation which ended up providing us with excellent labels on what was ultimately a difficult task on noisy data. Centaur understood what the issues were quickly from the beginning and were able to help steer the project for us. And more importantly they were a real joy to work with—they always went out of their way to help problem solve for us. Thanks for a really great experience!

Gareth Jones
Machine Learning Team Lead
Feebris

Explore more labeling solutions

Get started with Centaur Labs today