We’ve all had days when our voice goes hoarse, whether it’s after a cold, a concert, or an overexcited night at the stadium. Normally, rest and tea are enough. But for some people, the hoarseness never fades.

That lingering rasp can be an early warning sign of laryngeal cancer. The prognosis is variable, ranging from 35% to 78% survival over five years when treated. But the key element is picking it up early.

Today, the gold standards for diagnosis are nasal endoscopy or biopsy. Both are invasive, uncomfortable, and only recommended after symptoms appear or a clear abnormality is found. That delay can give the cancer more time to grow. Now, researchers have shown that abnormalities linked to laryngeal cancer can be diagnosed with AI from voice recordings.

“Early, noninvasive identification of these lesions using voice as a biomarker may improve diagnostic access and outcomes. In this study, we analyzed data from the initial release of the Bridge2AI-Voice dataset to evaluate which acoustic features best distinguish laryngeal cancer and benign vocal fold lesions from other vocal pathologies and healthy voice function,” write the authors of a new study.

The ‘Bridge2AI-Voice’ project is carried out within the US National Institute of Health’s ‘Bridge to Artificial Intelligence’ (Bridge2AI) consortium, a nationwide endeavor to apply AI to complex biomedical challenges.

“We show that with this dataset we could use vocal biomarkers to distinguish voices from patients with vocal fold lesions from those without such lesions,” said Dr Phillip Jenkins, a postdoctoral fellow in clinical informatics at Oregon Health & Science University, and the study’s corresponding author.

Detecting Cancer by Voice

The idea is part of a growing movement in medicine: biomarkers hidden in everyday signals. Your gait can reveal Parkinson’s risk. Your cough can hint at COVID-19. Now, your voice might carry the earliest clues of cancer.

In the study, researchers analyzed 12,523 voice recordings of 306 participants from across North America. A minority of them were from patients with known laryngeal cancer, benign vocal fold lesions, and two other conditions of the voice box: spasmodic dysphonia and unilateral vocal fold paralysis.

The idea was to see whether the algorithm could tell between benign conditions and cancer.

The researchers focused on the acoustic parameters of the voice. The mean fundamental frequency (pitch), jitter, variation in pitch, shimmer, and the harmonic-to-noise ratio (a measure of the relation between harmonic and noise components of speech) were all analyzed.

Rather curiously, they didn’t find any informative acoustic features among women. For men, the most useful information was in the fundamental frequency and the harmonic-to-noise ratio.

The researchers found marked differences in the harmonic-to-noise ratio and fundamental frequency between men without any voice disorder, men with benign vocal fold lesions, and men with laryngeal cancer. While they didn’t find any informative acoustic features among women, it is possible that a larger dataset would reveal such differences.

“These findings suggest that HNR, particularly its variability, may hold promise as a voice based marker for early detection and monitoring of vocal fold lesions. Further research with larger, more diverse populations is needed to refine these features and validate their clinical utility,” the study reads.

“Our results suggest that ethically sourced, large, multi‑institutional datasets like Bridge2AI‑Voice could soon help make our voice a practical biomarker for cancer risk in clinical care,” said Jenkins.

For Now, Just a Proof of Concept

For now, the method isn’t ready to go out into the world. This is a proof-of-principle, and the next step is to use the algorithms on more data and clinical settings.

For this, however, researchers need to gather more data, particularly on women.

“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. We then need to test the system to make sure it works equally well for women and men,” said Jenkins.

“Voice-based health tools are already being piloted. 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,” predicted Jenkins.

If refined and validated, such a tool could make early screening vastly easier. A short voice recording could become part of a routine check-up or even a smartphone app, flagging patients for further examination before symptoms become severe.

For laryngeal cancer, that could mean catching it not after months of hoarseness, but after a few seconds of speaking.

The study has been published in the journal Frontiers in Digital Health.