Published on
ADVEReadNOWISEMENT
A simple voice recording could one day help doctors spot early signs of throat cancer, according to new research.
In a study published in Frontiers in Digital Health, scientists found that artificial intelligence (AI) could potentially detect abnormal growths on the vocal cords, from benign nodules to early-stage laryngeal cancer, by analysing short voice recordings.
The findings could support efforts to find an easier, faster way to diagnose cancerous lesions on the vocal cords, also known as folds.
“With this dataset we could use vocal biomarkers to distinguish voices from patients with vocal fold lesions from those without such lesions,” said Phillip Jenkins, the study’s lead author and a postdoctoral researcher in clinical informatics at Oregon Health and Science University in the United States.
Why early detection of throat cancer matters
Cancer of the voice box, or larynx, affects more than a million people worldwide and kills roughly 100,000 every year. It is the 20th most common cancer in the world.
Smoking, alcohol use, and certain strains of HPV (human papillomavirus) are key risk factors, and survival rates vary from around 35 per cent to 90 per cent depending on how early the disease is diagnosed, according to Cancer Research UK.
One of the most common warning signs for laryngeal cancer is hoarseness or changes in the voice that last more than three weeks. Other symptoms include a persistent sore throat or cough, difficulty or pain when swallowing, a lump in the neck or throat, and ear pain.
Early detection of laryngeal cancer is crucial because it significantly improves survival rates and treatment outcomes.
Yet current diagnostic methods, including nasal endoscopies and biopsies, are invasive, uncomfortable, and often slow, requiring specialist equipment and expertise that many patients struggle to access quickly.
Developing a simple tool to flag early signs of vocal fold abnormalities through a quick voice recording could transform how throat cancer is detected – making it faster, more affordable and accessible to a wider population.
The next steps for AI-driven diagnosis
The research team examined about 12,500 voice recordings from 306 people across North America. They looked for subtle acoustic patterns, such as changes in pitch, loudness, and harmonic clarity.
The team identified clear differences for men in the harmonic-to-noise ratio and pitch between those with healthy voices, benign lesions, and cancer. No significant patterns were found in women, but the researchers say this may be due to the smaller dataset.
Jenkins said that the results indicate large datasets “could soon help make our voice a practical biomarker for cancer risk in clinical care”.
The next step is to train AI models on larger, professionally labelled datasets and test them in clinical settings. The team would also need to test the system to make sure it works well for both men and women, he said.
“Voice-based health tools are already being piloted,” Jenkins said.
“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”.