AI may diagnose voice box cancer just by listening to your voice

Your voice might reveal more than just what you say. It could hold hidden clues about benign conditions that you carry, like nodules or polyps, and even early stages of voice box cancer.

A recent study found that our voice could signal abnormalities in the vocal cords, potentially helping detect issues early.

AI detects early voice box cancer

The researchers developed an AI tool that could record the voice, analyze it, and predict the early signs of laryngeal cancer.

“Here 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, the author of the study and a postdoctoral fellow in clinical informatics at Oregon Health & Science University (OHSU).

Dr. Jenkins and his colleagues are part of the “Bridge2AI-Voice” project, and have been exploring the applications of AI to solve biomedical challenges. They studied 12,523 voice recordings from 306 participants across North America.

Understanding how the voice is produced and how cancer alters it helps explain why AI can detect problems early.

Understanding the voice box

The larynx, commonly known as the voice box, is a hollow tube situated in the middle of the neck that helps us make sounds including whispering, singing, or shouting. When we make sound, the vocal cords, muscle bands found in the larynx, vibrate and produce voice. 

Laryngeal cancer, commonly known as voice box cancer, poses a growing public health concern. It can be caused by certain types of human papillomavirus (HPV) and is also strongly linked to the frequent use of tobacco and alcohol.

Epidemiological studies suggest that in 2021 alone, doctors diagnosed around 1.1 million laryngeal cancer cases worldwide, and approximately 100,000 people died from it. 

If it is detected early and the tumor is located in a favorable spot, the chances of survival are high. If detected at an advanced stage, however, treatment can be more challenging.

Limitations of existing diagnostic methods

Currently, biopsies and video nasal endoscopies are used to diagnose voice box anomalies. 

A biopsy involves taking a small piece of tissue from the tumor and examining it under a microscope. It is an invasive procedure. 

A nasal endoscopy involves inserting a thin, flexible tube with a camera, called an endoscope, through the nasal passage to view the larynx.

While this is minimally invasive and offers a direct view of the affected site, it can still cause discomfort, pain, and in some cases, bleeding or fainting.

Since not all doctors can perform these diagnostic procedures, patients often face delays in searching for the right specialists for diagnosis. 

These drawbacks call for a safer, less invasive, and more accessible diagnostic approach, like the recently developed AI tool.

Measuring the features of the voice

The team measured the acoustic features – the physical characteristics – of the voice, such as quality, pitch, and loudness.  

Participants included individuals with diagnosed laryngeal cancer, benign vocal cord lesions, or other voice disorders. The researchers analyzed the Harmonic-to-noise ratio (HNR), which measures sound clarity. 

For instance, if you have a high HNR, your sound is clear, with clear harmonics and less noise. People with laryngeal anomalies, on the other hand, tend to have a low HNR.

They also watched the fundamental frequency closely. This refers to the average pitch of the voice.  By comparing these measurements across participants, the team identified distinct patterns that were linked to vocal health.

What the study reveals

The study found clear differences in the voice parameters among men with no voice disorder, those with benign vocal cord lesions, and those with laryngeal cancer.  

They did not spot clear differences in these parameters among women participants, but hope that a larger dataset could reveal them.

The authors concluded that this approach could clinically evaluate vocal fold lesions and early laryngeal cancer, at least in men. 

“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 Dr. Jenkins.

Next step for AI voice tools

Since the authors showed that AI can be effective in the early diagnosis of laryngeal cancer, the next step is to test it with more voice samples in clinical settings. 

To turn this into a useful AI tool, the researchers need to analyze more voice recordings using the same approach. “We then need to test the system to make sure it works equally well for women and men,” said Dr. Jenkins.

According to the study, AI health tools for voice analysis are still not ready for use in clinics or hospitals. However, they could move into early-stage testing within the next couple of years. 

The AI tool could offer a promising addition to the current diagnostic methods.

The full study was published in the journal Frontiers in Digital Health.

—–

Like what you read? Subscribe to our newsletter for engaging articles, exclusive content, and the latest updates. 

Check us out on EarthSnap, a free app brought to you by Eric Ralls and Earth.com.

—–