A group of American researchers demonstrated that using the patient’s voice, artificial intelligence (AI) can assist in the early detection of laryngeal or voice box cancer.

Voice box cancer is a significant public health concern. An estimated 1.1 million instances of laryngeal cancer were reported globally in 2021, and the disease claimed the lives of about 100,000 people.

Human papillomavirus infection, alcohol misuse, and smoking are risk factors.

Depending on the tumor’s stage and location within the voice box, the prognosis for laryngeal cancer varies from 35% to 78% survival over five years when treated.

Researchers at Oregon Health & Science University have now demonstrated that artificial intelligence (AI) can identify anomalies of the vocal folds from speech sounds.

Similar to nodules or polyps, these “vocal fold lesions” can be benign, but they can potentially be the first signs of laryngeal cancer.

The scientists stated in the paper published in the journal Frontiers in Digital Health that these proof-of-principle results pave the way for a new use of AI, specifically the recognition of the early warning phases of laryngeal cancer from voice recordings.

Dr. Phillip Jenkins, a postdoctoral researcher in clinical informatics at Oregon, stated, “Here we demonstrate that we could use vocal biomarkers to distinguish voices from patients with vocal fold lesions from those without such lesions.”

Using 12,523 voice recordings of 306 people from North America, Jenkins and his team examined differences in tone, pitch, loudness, and clarity.

A small percentage came from individuals who had unilateral vocal fold paralysis, spasmodic dysphonia, benign vocal fold lesions, or known laryngeal malignancy.

The mean fundamental frequency (pitch), jitter (pitch variation within speech), shimmer (amplitude variation), and the harmonic-to-noise ratio (a measure of the relationship between harmonic and noise components of speech) were among the acoustic characteristics of the voice that the researchers concentrated on.

They discovered significant variations in the fundamental frequency and harmonic-to-noise ratio among men with laryngeal cancer, individuals with benign vocal fold lesions, and men without any voice disorders.

Although a larger dataset would show such disparities, they were unable to uncover any meaningful acoustic patterns among women.

According to the researchers, variation in the harmonic-to-noise ratio can be useful for early detection of laryngeal cancer, at least in men, and for tracking the clinical progression of vocal fold lesions.