Algorithm predicts heart disease based on audio recordings

Researchers have developed an algorithm that analyzes audio recordings to predict the occurrence of heart conditions. coronary artery diseasea result of blockage of the blood vessels that carry blood to the heart.

Technology by examining the features such as frequency, amplitude, pitch and cadence of the recordings, 2.6 times Mayr have serious problems are associated with disease and are three times more likely to develop plaque in the arteries identified by medical testing.

The study’s lead author, Jaskanwal Deep Singh Sara, suggests that in the future voice technology could complement existing medical studies and healthcare modalities: “Providing a voice sample can become very intuitive, and even pleasing and scalable for patients. It’s a way for us to improve patient management.”

How was the algorithm developed?

Based on artificial intelligence, the algorithm was trained to analyze more than eighty properties of sound recordings, such as frequency and tone. For this purpose, more 10 thousand samples Your voice gathered in Israel

It was used in a study that asked 108 participants to record three 30-second audio samples each using a mobile phone app: the first was a text-to-speech; in the second, describing a positive experience; third, a negative experience. The algorithm rated the user on a scale of -1 to 1 after analyzing the recordings.

Participants were followed by researchers for two years. Of those with the highest score, 58% were admitted to the hospital with chest pain or acute coronary syndrome.this was true for about 30% of respondents with lower scores.

Sara, this phenomenon autonomic nervous systemregulating some speech mechanisms and cardiovascular system. Thus, the sound can indicate the functioning of this system and therefore the cardiovascular system.

“We ourselves cannot hear about these particular features. This technology uses machine learning using our human brains and ears to measure something that cannot be easily measured for us,” explains Sara.

He suggests that in the future, voice analytics could help identify individuals who would benefit from closer monitoring of heart disease-related events.

The study was presented April 2 at the 71st Annual Scientific Session of the American College of Cardiology and is simultaneously published in the journal Science. Mayo Clinic Proceedings.