AI May Soon Detect Cancer Just by Listening to You Speak

New research explores how hidden patterns in the human voice could serve as early indicators of disease.

Cancer of the larynx, often called the voice box, remains a major global health concern. In 2021, about 1.1 million people were diagnosed worldwide, and roughly 100,000 died from the disease. Smoking, heavy alcohol use, and infection with human papillomavirus are key risk factors. Survival rates vary widely, ranging from 35% to 78% over five years with treatment, depending on where the tumor develops and how advanced it is at diagnosis.

Early detection plays a critical role in improving outcomes. Today, diagnosis typically relies on video nasal endoscopy and tissue biopsies, which are invasive and can be difficult to access quickly. Delays in seeing a specialist may slow diagnosis and treatment.

New research published in Frontiers in Digital Health suggests a different approach. Scientists found that subtle changes in a person’s voice can reveal abnormalities in the vocal folds. These “vocal fold lesions” may be harmless, such as nodules or polyps, but they can also signal early-stage laryngeal cancer. The findings point to a potential new use for artificial intelligence: identifying early warning signs of cancer through voice analysis.

“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, a postdoctoral fellow in clinical informatics at Oregon Health & Science University, and the study’s corresponding author.

Voice messages

Jenkins and his team are part of the ‘Bridge2AI-Voice’ project within the US National Institute of Health’s ‘Bridge to Artificial Intelligence’ (Bridge2AI) consortium. This nationwide effort aims to apply AI to complex biomedical problems. For this study, the researchers examined tone, pitch, volume, and clarity using the first public release of the Bridge2AI-Voice dataset, which includes 12,523 recordings from 306 participants across North America.

Only a portion of these recordings came from people with diagnosed laryngeal cancer, benign vocal fold lesions, or other voice disorders such as spasmodic dysphonia and unilateral vocal fold paralysis.

The team analyzed several measurable features of speech. These included mean fundamental frequency, or pitch, along with jitter, which reflects small variations in pitch, and shimmer, which captures changes in amplitude. They also measured the harmonic-to-noise ratio, which compares structured sound to background noise in speech.

Clear differences emerged in the harmonic-to-noise ratio and pitch among men without voice disorders, men with benign lesions, and men with laryngeal cancer. Similar patterns were not identified in women, although the researchers note that a larger dataset may reveal meaningful trends.

The study suggests that changes in the harmonic-to-noise ratio may help track how vocal fold lesions develop and could support early detection of laryngeal cancer, particularly in men.

“Our results suggest that ethically sourced, large, multi-institutional datasets like Bridge2AI-Voice could soon help make our voice a practical biomarker for cancer risk in clinical care,” said Jenkins.

Building a bridge to AI

With these initial results in place, the next step is to apply the algorithms to larger datasets and evaluate their performance in clinical settings.

“To move from this study to an AI tool that recognizes vocal fold lesions, we would train models using an even larger dataset of voice recordings, labeled by professionals. We then need to test the system to make sure it works equally well for women and men,” said Jenkins.

“Voice-based health tools are already being piloted. Building on our findings, I estimate that with larger datasets and clinical validation, similar tools to detect vocal fold lesions might enter pilot testing in the next couple of years,” predicted Jenkins.

Reference: “Voice as a biomarker: exploratory analysis for benign and malignant vocal fold lesions” by Phillip Jenkins, Rylan Harrison, Steven Bedrick, Lisa Karstens, Bridge2AI-Voice Consortium , William Hersh, Yael Bensoussan, Olivier Elemento, Anais Rameau, Alexandros Sigaras, Satrajit Ghosh, Maria Powell, Vardit Ravitsky, Jean Christophe Belisle-Pipon, David Dorr, Phillip Payne, Alistair Johnson, Ruth Bahr, Donald Bolser, Frank Rudzicz, Jordan Lerner-Ellis, Kathy Jenkins, Shaheen Awan, Micah Boyer, William Hersh, Andrea Krussel, Steven Bedrick, Toufeeq Ahmed Syed, Jamie Toghranegar, James Anibal, Duncan Sutherland, Enrique Diaz-Ocampo, Elizabeth Silberhoz, John Costello, Alexander Gelbard, Kimberly Vinson, Tempestt Neal, Lochana Jayachandran, Evan Ng, Selina Casalino, Yassmeen Abdel-Aty, Karim Hanna, Theresa Zesiewicz, Elijah Moothedan, Emily Evangelista, Samantha Salvi Cruz, Robin Zhao, Mohamed Ebraheem, Karlee Newberry, Iris De Santiago, Ellie Eiseman, JM Rahman, Stacy Jo and Anna Goldenberg, 25 June 2025, Frontiers in Digital Health.
DOI: 10.3389/fdgth.2025.1609811

Never miss a breakthrough: Join the SciTechDaily newsletter.
Follow us on Google and Google News.