• Posted August 13, 2025

AI Uses Voice To Detect Throat Cancer

A person’s own voice might soon be a means of detecting whether they’re suffering throat cancer, a new study says.

Men with cancer of the larynx, or voice box, have distinct differences in their voices that could be detected with trained artificial intelligence (AI), researchers reported Aug. 12 in the journal Frontiers in Digital Health.

These differences are caused by potentially cancerous lesions that have cropped up in a person’s vocal folds – the two bands of muscle tissue in the larynx that produce sound, also known as vocal cords.

“We could use vocal biomarkers to distinguish voices from patients with vocal fold lesions from those without such lesions,” lead researcher Dr. Phillip Jenkins, a postdoctoral fellow in clinical informatics at Oregon Health & Science University in Portland, said in a news release.

Catching voice box cancer early can be a matter of life or death.

There were an estimated 1.1 million cases of laryngeal cancer worldwide in 2021, and about 100,000 people died from it, researchers said in background notes. Risk factors include smoking, drinking and HPV infection.

A person’s odds of five-year survival can be as high as 78% if their throat cancer is caught at an early stage, or as low as 35% if it’s caught late, researchers said. 

For the study, researchers analyzed more than 12,500 voice recordings from 306 people across North America. These included a handful of people with either laryngeal cancer, benign vocal cord lesions or other vocal disorders.

Researchers discovered that the voices of men with laryngeal cancer exhibited marked differences in harmonic-to-noise ratio, which judges the amount of noise in a person’s speech.

Men with laryngeal cancer also showed differences in the pitch of their voices, results show.

The team concluded that harmonic-to-noise ratio in particular might be used to track vocal cord lesions and potentially detect voice box cancer at an early stage, at least in men.

They weren’t able to detect any differences among women with laryngeal cancer, but are hopeful a larger dataset might reveal such differences.

The next step will be to feed the AI more data and test its effectiveness with patients in clinical settings, researchers said.

“To move from this study to an AI tool that recognizes vocal fold lesions, we would train models using an even larger dataset of voice recordings, labeled by professionals,” Jenkins said. Then, the system will need to be tested to make sure it works equally well for both women and men.

“Voice-based health tools are already being piloted,” Jenkins added. “Building on our findings, I estimate that with larger datasets and clinical validation, similar tools to detect vocal fold lesions might enter pilot testing in the next couple of years.”

More information

The American Cancer Society has more on throat cancers.

SOURCE: Frontiers, news release, Aug. 12, 2025