Particle.news

Download on the App Store

AI Voice Analysis Differentiates Benign and Cancerous Vocal-Fold Lesions in Men

The proof-of-principle study harnesses voice biomarkers to flag male vocal-fold lesions, underscoring the need for larger, labeled datasets for rigorous AI validation.

Overview

  • Analysis of 12,523 voice recordings from 306 North American participants identified harmonic-to-noise ratio variation and pitch differences that distinguish healthy male voices from those with benign lesions and laryngeal cancer.
  • Case counts for men were modest—122 healthy, 13 benign lesion, and 10 cancer recordings—limiting statistical power and generalizability.
  • No statistically significant acoustic markers were detected among women, highlighting equity concerns and the need for more diverse voice data.
  • Researchers emphasize that these proof-of-principle results require training on larger, professionally labeled datasets and rigorous clinical testing before AI-based triage tools can be piloted.
  • Voice-based screening could offer a non-invasive alternative to endoscopy and biopsy, potentially speeding up detection of voice-box disease.