AI listens to toilet noises to guess if people have diarrhea

Artificial intelligence that can detect diarrhea with 98% accuracy from recordings of toilet noises could help track disease outbreaks, such as cholera

By Carissa Wong

An artificial intelligence can detect diarrhea with up to 98% accuracy by analyzing sounds emanating from the toilet. This skill could help us track disease outbreaks such as cholera.

Maia Gatlin at the Georgia Institute of Technology and colleagues collected 350 recordings of toilet sounds from YouTube and a sound database Soundsnap – covering standard defecation, diarrhoea, urination and flatulence.

The researchers then used 70% of the recordings to train an AI to recognize audible differences between the four types of excretion. Once they confirmed that the AI ​​could do this consistently with an additional 10% of the data, they tested the AI’s performance using the remaining 20% ​​of the records.

This revealed that the AI ​​could correctly classify a shedding event as diarrheal or non-diarrheal with 98% accuracy, if background noise – such as people talking – was filtered out, and with 96% accuracy if background noise was maintained.

Using this approach to track outbreaks would involve placing microphones near public restrooms and feeding the data to AI, Gatlin says.

The researchers created an installation that could be mounted in the toilet. A microphone picks up the noise of toilet use, which is recorded on a microprocessor in a nearby “Diarrhea Detector” box (pictured above) that incorporates a machine-learning AI model. The signal is processed and evaluated before being classified as diarrhea or not.

Diarrheal diseases, such as cholera, can lead to death if left untreated, and automatically detecting levels of diarrhea in the community could help track outbreaks and reduce the spread of disease.

However, the sound of excretion events depends on the type of toilet used.

“Many areas where cholera is rampant don’t have the same types of toilets that we have in the US or the UK, so we would need to develop an AI for the sounds made in different types of toilets,” explains Gatlin, who announced the results at a Meeting of the Acoustical Society of America December 5.

Using online recordings to develop the AI ​​also meant that researchers had to manually listen to the recordings and decide whether the audio tags accurately described the type of shedding event – ​​without knowing for sure what type was being recorded.

“In the future, we would like to collect real-world excretion records and develop AI on those,” says Gatlin.

Sign up for our free Checkup newsletter that gives you the health, food and fitness news you can trust, every Saturday