Can AI Identify Early Laryngeal Cancer from Voice Sounds?

New Delhi, Aug 12 (NationPress) A group of scientists from the United States has demonstrated that Artificial Intelligence (AI) can assist in identifying early stages of laryngeal or voice box cancer through the sounds produced by a patient's voice.

Laryngeal cancer poses a significant public health challenge. In 2021, it was estimated that there were approximately 1.1 million cases of this cancer globally, leading to nearly 100,000 deaths.

Common risk factors include smoking, alcohol consumption, and infection with human papillomavirus.

The survival rate for laryngeal cancer varies significantly, ranging from 35% to 78% over five years after treatment, depending on the stage and location of the tumor.

Recently, researchers from Oregon Health & Science University have shown that abnormalities in the vocal folds can be identified through voice analysis using AI.

These abnormalities, termed 'vocal fold lesions,' can be non-cancerous, such as nodules or polyps, or may signify the early onset of laryngeal cancer.

The findings from this study pave the way for a novel application of AI: recognizing the preliminary warning signs of laryngeal cancer using voice recordings, as reported in the journal Frontiers in Digital Health.

“Our research indicates that we can leverage vocal biomarkers to differentiate between voices of patients with vocal fold lesions and those without,” stated Dr. Phillip Jenkins, a postdoctoral fellow in clinical informatics at Oregon.

In their investigation, Jenkins and colleagues analyzed tonal variations, including pitch, volume, and clarity, utilizing 12,523 voice recordings from 306 participants across North America.

A small subset of participants had known laryngeal cancer, benign vocal fold lesions, or two other voice box conditions: spasmodic dysphonia and unilateral vocal fold paralysis.

The research focused on various acoustic characteristics of the voice, such as mean fundamental frequency (pitch), jitter (pitch variation during speech), shimmer (amplitude variation), and the harmonic-to-noise ratio, which assesses the balance between harmonic and noise elements in speech.

Significant differences were observed in the harmonic-to-noise ratio and fundamental frequency among men without voice disorders, those with benign vocal fold lesions, and those with laryngeal cancer.

No informative acoustic characteristics were identified among women; however, it is plausible that a larger dataset might unveil such distinctions.

Variations in the harmonic-to-noise ratio could aid in monitoring the clinical progression of vocal fold lesions and in early detection of laryngeal cancer, particularly in men, according to the researchers.