Butterfly Network has received the US Food and Drug Administration 510(k) approval for its AI-enabled Auto B-line Counter to help evaluate adults with suspected abnormal lung conditions.

The Auto B-line Counter is designed to deploy deep learning technology to produce a B-line count from just a six-second ultrasound clip.

B-lines, also known as ultrasound lung comets, appear as bright, vertical lines on an ultrasound scan, indicating wetness in the lungs and are associated with pulmonary air-space disease.

Unlike traditionally manual, subjective counting processes, the device enables more consistent interpretation of B-lines, said the US digital health company.

Butterfly Network founder and interim chief executive officer Jonathan Rothberg said: “Our goal at Butterfly is to give healthcare practitioners, and eventually consumers, a real-time full colour, annotated, window into the human body.

“Applying AI to make ultrasound easier to use is core to Butterfly and will enable powerful ultrasound to be in the palm of more clinician’s hands, across specialities, to monitor, assess, and prescribe treatments in a more informed way.

“Our AI-enabled Auto B-line Counter empowers providers to assess lung conditions faster and with more confidence – and in turn, will aid in earlier detection, diagnosis, and treatment of cardiovascular diseases, a leading cause of death globally, taking nearly 18 million lives each year.”

Butterfly said that its Auto B-line Counter algorithm uses the advanced instant percent counting method, which is said to be more reliable than incumbent individual line counting methods.

In the instant percent counting method, numbers are assigned confluent B-lines, by the percentage of rib space occupied in addition to counting discrete B-lines.

It enables the trained providers to simply place the probe and receive a reliable number count right on their screen, said the medical device maker.

Butterfly uses its secure Butterfly Cloud to access more than 3.5 million de-identified ultrasound images to develop and train its AI algorithms.

The data comes from various sites across the US, with a broad and diverse range of age, gender, body mass index, ethnicity, and race.