Researchers are exploring how Artificial Intelligence (AI) can detect subtle changes in a person’s voice to identify small lesions on the vocal cords, particularly in patients with laryngeal cancer. Experts believe AI could be trained to recognize these changes within the next few years.
In the United Kingdom, about 2,000 people are diagnosed with laryngeal cancer annually. A persistent voice change lasting more than three weeks is considered a warning sign and should prompt a medical examination.
AI could potentially replace current invasive diagnostic methods, such as biopsies or nasoendoscopy—a procedure in which a thin tube with a camera is inserted through the nose to examine the throat.
The Bridge2AI-Voice project in the United States collected voice samples from 300 people, including some diagnosed with laryngeal cancer. Researchers analyzed six key voice characteristics, including pitch, jitter (small variations in pitch), shimmer (variations in loudness), and the harmonic-to-noise ratio (the relationship between tonal quality and background noise).
The study found significant differences between men with vocal cancer or other vocal disorders—such as lesions or nodules—and those without. The harmonic-to-noise ratio showed the most notable variation. Researchers believe similar patterns could be observed among women once additional data is collected.
Dr. Phillip Jenkins from Oregon Health & Science University said, “To turn this research into an AI tool capable of identifying vocal cord lesions, we need to train the system using a large dataset of voice samples that have been carefully analyzed by medical experts.”
The team plans to test the AI system to ensure accuracy for both men and women. Voice-based health assessment tools are already in experimental stages, and once validated, they could enter clinical trials within the next two years.
