AI Technology Aims to Detect Early Voice Box Cancer Through Voice Analysis

Voice box or laryngeal cancer remains a significant public health challenge globally. In 2021, approximately 1.1 million individuals were diagnosed with laryngeal cancer worldwide, resulting in around 100,000 deaths. Major risk factors include tobacco smoking, excessive alcohol consumption, and infections such as human papillomavirus (HPV). The five-year survival rate varies between 35% and 78%, depending on the stage of the tumor and its precise location within the voice box.

Early detection is crucial for improving patient outcomes, yet current diagnostic procedures are invasive and require specialized equipment and expertise. Standard diagnosis involves video nasal endoscopy paired with biopsies, which can be time-consuming and uncomfortable.

Innovative research published in Frontiers in Digital Health explores how artificial intelligence (AI) can transform this process. Researchers have demonstrated that abnormalities associated with vocal fold lesions—some benign, like nodules or polyps, and others potentially malignant—can be identified from voice recordings. This approach offers a non-invasive, accessible way to screen for early signs of laryngeal cancer.

Led by Dr. Phillip Jenkins of Oregon Health & Science University, the research team analyzed voice samples collected from a diverse group of over 300 participants across North America, part of the NIH-supported Bridge2AI-Voice project. They examined various acoustic features such as pitch, jitter, shimmer, and noise ratio to distinguish between healthy voices and those with vocal fold lesions or cancer.

The study found significant differences in the harmonic-to-noise ratio and fundamental frequency, particularly in male participants, suggesting these metrics could serve as early biomarkers for the disease. Although women’s voices did not show clear differences in this initial study, larger datasets may reveal similar patterns.

This promising research opens pathways for the development of AI-powered tools that could assist clinicians in early detection of voice box cancers. Future efforts involve training more advanced models on expanded datasets and validating their effectiveness in clinical environments. Such tools could drastically reduce the time and invasiveness of diagnosis, ultimately saving lives.

Building on this proof of concept, researchers anticipate that with further refinement and validation, voice-based screening tools could be integrated into routine health checks within a few years, offering a new frontline in cancer prevention and early intervention.

Source: https://medicalxpress.com/news/2025-08-ai-early-voice-cancer.html