Artificial intelligence could transform how doctors diagnose cancer of the voice box by analysing recordings of patients' voices, a new study suggests.
Scientists have discovered that lesions on the vocal folds and early-stage laryngeal cancer create distinct acoustic patterns that AI systems can identify.
As a result, the technology could offer a non-invasive alternative to current diagnostic methods within the next few years.
Researchers from Oregon Health & Science University examined over 12,500 voice recordings from 306 participants across North America as part of the Bridge2AI-Voice project.
The sound of someone's voice could be an early warning sign of throat cancer
|GETTY
Their findings suggest voice analysis could become a practical screening tool for detecting potentially cancerous changes.
The study revealed significant acoustic variations between healthy voices and those affected by vocal fold lesions or cancer, particularly among male participants.
Key differences emerged in pitch levels and the harmonic-to-noise ratio, which measures the relationship between harmonic and noise elements in speech.
"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 postdoctoral fellow Dr Phillip Jenkins.
Although the research identified clear patterns in men's voices, no comparable acoustic markers were detected in women, but researchers have suggested that a more extensive dataset might reveal such differences.
LATEST DEVELOPMENTS
Laryngeal cancer affects approximately 1.1 million people globally each year, with around 100,000 deaths recorded in 2021 alone.
Survival rates vary between 35 and 78 per cent over five years following treatment, depending on tumour stage and location.
Currently, diagnosis of the disease requires video nasal endoscopy and biopsies - procedures that are both invasive and time-consuming.
Additionally, delays in reaching specialists who can perform these examinations often postpone diagnosis, potentially affecting patient outcomes.
The new AI approach could eliminate these barriers by detecting abnormalities through simple voice recordings, offering faster and more accessible screening for at-risk patients.
Voice analysis could become a non-invasive screening tool
|GETTY
The researchers believe voice-based diagnostic tools could undergo pilot testing within two years, pending further development.
"Our results suggest that ethically sourced, large, multi-institutional datasets like Bridge2AIVoice could soon help make our voice a practical biomarker for cancer risk in clinical care," the fellow explained.
Before clinical implementation, however, scientists must expand their datasets and ensure the technology performs equally well across all demographics.
"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, labelled by professionals," Jenkins explained. "We then need to test the system to make sure it works equally well for women and men."