AI could soon be able to tell whether patients have cancer of the voice box using just a voice note, according to new research.
Scientists recorded the voices of men with and without abormalities in their vocal folds – which can be an early sign of laryngeal cancer – and found differences in vocal qualities including pitch, volume, and clarity. They now say AI could be used to detect these “vocal biomarkers”, leading to earlier, less invasive diagnosis.
Researchers at Oregon Health and Science University believe voice notes could now be used to train an AI tool that recognises vocal fold lesions.
Using 12,523 voice recordings from 306 participants across North America, they found distinctive vocal differences in men suffering from laryngeal cancer, men with vocal fold lesions, and men with healthy vocal folds. However, researchers said similar hallmark differences were not detected in women.
They are now hoping to collect more recordings of people with and without the distinctive vocal fold lesions to create a bigger dataset for tools to work from.
In the UK, there are more than 2,000 new cases of laryngeal cancer each year. Symptoms can include a change in your voice, such as sounding hoarse, a high-pitched wheezing noise when you breathe, and a long-lasting cough.
“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 study’s corresponding author said.
“To move from this study to an AI tool that recognises 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.
“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,” he predicted.
It comes after research from US-based Klick Labs, which created an AI model capable of distinguishing whether a person has Type 2 diabetes from six to 10 seconds of voice audio. The study involved analysing 18,000 recordings in order to identify acoustic features that differentiated non diabetics from diabetics and reported an 89 per cent accuracy rating for women and 86 per cent for men.
Jaycee Kaufman, a research scientist at Klick Labs, praised the future potential for AI-powered voice tools in healthcare, saying: “Current methods of detection can require a lot of time, travel and cost. Voice technology has the potential to remove these barriers entirely.”
