AI Can Detect Early Voice Box Cancer By Analyzing Subtle Changes in Voice

In a recent development, scientists from the US have demonstrated that Artificial Intelligence (AI) can detect early-stage laryngeal, or voice box, cancer by analyzing subtle changes in a patient’s voice. The findings, published in Frontiers in Digital Health, point to the potential of using voice recordings as a non-invasive screening tool for early diagnosis.

Laryngeal cancer poses a significant global health challenge, with approximately 1.1 million cases recorded worldwide in 2021 and around 100,000 deaths. Common risk factors include tobacco use, excessive alcohol consumption, and human papillomavirus (HPV) infection. Depending on the cancer’s stage and location within the larynx, five-year survival rates can range from 35% to 78% if treated.

The research team from Oregon Health & Science University focused on detecting abnormalities in the vocal folds—small lesions that may be benign, like polyps or nodules, or indicative of early-stage cancer—using AI tools trained on voice data.

“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, postdoctoral fellow in clinical informatics at Oregon and lead author of the study.

The team analyzed over 12,500 voice recordings from 306 participants across North America, including patients diagnosed with laryngeal cancer, benign lesions, or other voice disorders such as spasmodic dysphonia and unilateral vocal fold paralysis.

Researchers assessed several vocal characteristics—such as pitch (mean fundamental frequency), jitter (pitch variation), shimmer (amplitude variation), and harmonic-to-noise ratio (HNR), which indicates the balance between normal and irregular sound components.

Notably, significant differences were observed in HNR and pitch among men without voice disorders, men with benign lesions, and those with cancer. These vocal markers could be valuable in tracking disease progression and identifying early signs of cancer. However, similar patterns were not observed in women, possibly due to the smaller dataset—something future research could address.

The study paves the way for AI-powered voice analysis as a promising early detection tool, particularly for male patients at risk of laryngeal cancer.