Your voice may reveal more than just what you say. It can hold hidden clues about the benign state you carry, as well as the early stages of voice box cancer, like nodules and polyps.
Recent research has shown that our voices may show abnormalities in the vocal cords, which may help to detect problems early.
AI detects early voice box cancer
Researchers have developed an AI tool that can record, analyze voices and predict early signs of laryngeal cancer.
“Here, this dataset shows that vocal biomarkers can be used to distinguish voices from patients with voice lesions without such lesions,” says Dr. Philip Jenkins, the author of the study and a postdoctoral researcher in clinical informatics at Oregon Health & Science University (OHSU).
Dr. Jenkins and his colleagues are part of the Bridge2ai-Voice project, examining the applications of AI to solve biomedical challenges. They studied 12,523 audio recordings from 306 participants in North America.
Understanding how voices are generated and how cancer changes can help explain why AI can detect problems early.
Understand the voice box
The larynx, commonly known as the voice box, is a hollow tube located in the center of the neck that helps to create sounds such as whispers, singing, and screaming. When you make a sound, it vibrates the vocal cords, muscle bands and voices found in the larynx, creating a voice.
Laryngeal cancer, commonly known as voice box cancer, is growing public health concerns. It can be caused by certain types of human papillomaviruses (HPV) and is also strongly associated with the frequent use of tobacco and alcohol.
Epidemiological studies suggest that in 2021 alone, doctors diagnosed around 1.1 million cases of laryngeal cancer worldwide, with around 100,000 deaths.
If it is detected early and the tumor is in the preferred location, there is a high chance of survival. However, treatment can be more difficult if detected at a sophisticated stage.
Limitations to existing diagnostic methods
Currently, biopsies and video nasal endoscopy are used to diagnose abnormalities in the audio box.
A biopsy involves taking small amounts of tissue from the tumor and examining it under a microscope. It's an invasive procedure.
Nasal endoscopy involves inserting a thin, flexible tube into a camera called an endoscope through the nasal scoop to see the larynx.
This is minimally invasive and allows you to see the affected site directly, but can cause discomfort, pain, and in some cases bleeding or fainting.
Not all physicians can carry out these diagnostic procedures, and patients often face delays in searching for the right expert for a diagnosis.
These drawbacks require a more secure, less invasive and more accessible diagnostic approach, like the recently developed AI tools.
Measurement of speech features
The team measured the acoustic features (physical properties) of the voice, including quality, pitch and loudness.
Participants include individuals diagnosed with laryngeal cancer, benign vocal cord lesions, or other vocal dysfunction. The researchers analyzed harmonics and noise ratios (HNR) that measure sound clarity.
For example, if the HNR is high, the sound is clear, with clear harmonics and less noise. On the other hand, people with laryngeal abnormalities tend to have a lower HNR.
They also carefully looked at the basic frequency. This refers to the average pitch of the voice. By comparing these measures between participants, the team identified different patterns linked to vocal health.
What this study reveals
This study found clear differences in speech parameters in men without speech disorders, men with benign vocal cord lesions, and patients with laryngeal cancer.
They found no clear differences in these parameters among female participants, but they hope that a larger dataset will reveal them.
The authors concluded that this approach can clinically evaluate vocal cord lesions and early laryngeal cancer, at least in men.
“Our results suggest that large, ethically sourced, multicenter datasets like Bridge2AI-Voice will soon help our voices become practical biomarkers for cancer risk in clinical care,” says Dr. Jenkins.
The authors showed that AI is effective in early diagnosis of laryngeal cancer, so the next step is to test it with more audio samples in a clinical setting.
To turn this into a useful AI tool, researchers will need to use the same approach to analyze more audio recordings. “The next thing we need to do is test the system to make sure it works just as well for women and men,” Dr. Jenkins said.
According to the study, AI health tools for speech analysis have not yet been available for use in clinics or hospitals. However, they could move to early stage testing within the next few years.
AI tools can provide promising additions to current diagnostic methods.
The complete study was published in the journal Digital Health Frontier.
– –
Like what you read? Subscribe to our newsletter for compelling articles, exclusive content and the latest updates.
Check out us on EarthSnap, a free app provided by Eric Ralls and Earth.com.
– –
