An exploratory study published in Frontiers in Digital Health suggests that subtle variations in voice patterns, particularly fluctuations in the harmonic-to-noise ratio (HNR), could serve as an early biomarker for vocal fold lesions. Researchers from Oregon Health & Science University and Portland State University analyzed 12,523 voice recordings from 306 individuals within the NIH Bridge2AI-Voice dataset to evaluate whether acoustic features could distinguish benign or malignant lesions from healthy voices and other disorders.
The team found that mean HNR and fundamental frequency differed between individuals with benign lesions and those without vocal disorders, while HNR variability helped separate benign from malignant lesions. However, these distinctions were not significant in female participants, possibly due to limited sample size. Jitter and shimmer showed no clear associations.
The findings indicate that HNR variability may reflect the consistency of vocal fold vibration and could be useful for monitoring lesion progression or identifying laryngeal cancer at an earlier stage. However, the study did not find clear acoustic markers to differentiate lesions from other voice disorders such as spasmodic dysphonia or vocal fold paralysis.
Researchers emphasize that these results are preliminary and do not yet constitute a clinical screening tool. Larger, sex-balanced cohorts and additional acoustic analyses are needed to confirm whether voice-based AI systems could help identify patients with subtle vocal changes who might otherwise delay medical evaluation.
