Could Your Smartphone Detect Signs Of Cancer From The Sound Of Your Voice?

(© jittawit.21 – stock.adobe.com)

PORTLAND, Ore. — You’re reading aloud to your child, chatting with a colleague, or ordering coffee. Your voice sounds normal to you, but hidden in those everyday sounds could be clues about your throat health that doctors might miss during a routine checkup.

A new exploratory study suggests the human voice may carry acoustic patterns linked to vocal fold lesions, including laryngeal cancer, but clinical screening tools will require much larger, validated studies before becoming reality.

Researchers from Oregon Health & Science University found that certain voice characteristics differ between healthy voices and those with vocal fold lesions. The research, published in Frontiers in Digital Health, points toward a future where analyzing your voice could become a quick screening step, especially valuable for people in rural areas who can’t easily reach specialized throat doctors.

How Common Are Vocal Fold Lesions—and How Are They Diagnosed Today?

In 2017, there were 13,150 reported cases of laryngeal cancer and 3,710 deaths, according to StatPearls, a U.S.-based medical reference. Meanwhile, a population-based study estimated 12.47% prevalence of benign vocal fold lesions in its cohort. These conditions can cause serious problems for teachers, singers, salespeople, and anyone whose career depends on clear speech.

Current diagnosis requires expensive equipment and visits to specialized voice clinics that many rural communities simply don’t have. Patients must undergo laryngoscopy, a procedure where doctors insert a scope through the nose or mouth to examine the vocal folds (also known as vocal cords) directly. By the time people notice voice changes, their conditions may have already progressed considerably.

“One of the first symptoms presented by patients with glottic organic lesions is dysphonia,” the researchers noted, referring to voice problems that make speaking difficult. Access to specialized care “can be limited outside of major urban centers with interdisciplinary voice clinics.”

Voice, speech, and respiratory sounds provide clinical insights into patients’ health status. In the age of artificial intelligence, patients’ audio recordings are being investigated as digital biomarkers for early detection of a broad range of conditions, including laryngeal pathology, neurological and psychological disorders, head and neck cancers, and diabetes.

How Scientists Tested Voice Biomarkers

Researchers analyzed recordings from 176 participants across five North American medical centers using the Bridge2AI-Voice dataset. Each person read the “Rainbow Passage,” a standard paragraph containing all English sounds that speech specialists use for voice evaluation.

The team examined four voice measurements: voice clarity measurements (called harmonic-to-noise ratio or HNR), jitter (pitch fluctuations), shimmer (loudness variations), and voice pitch (called fundamental frequency). Of these, HNR and its variability stood out. When vocal folds work normally, they create regular sound waves that produce clear speech. Lesions interfere with this process, introducing irregular noise and breathing sounds. The ratio between clean sound and background noise becomes a potential signature of vocal fold health.

Voice clarity measurements compare the ratio of regular vocal cord vibrations when speaking to background noise produced from turbulence as air flows through the glottis. A possible source of this turbulence is the improper closing of the vocal cords.

Results: Clearer Voice Patterns in Men, Less Evidence in Women

When researchers compared the voice recordings, they saw clear differences between people with healthy voices and those with benign growths on their vocal folds. These differences showed up most strongly in HNR measurements. They also found that overall pitch was different between these groups.

When benign lesions were compared to laryngeal cancer, the only statistically reliable difference was in the steadiness of that voice clarity measure (HNR variability).

For men, these distinctions were especially noticeable. Men with benign lesions stood out from both healthy men and those with cancer, mainly because their voices had more irregular background noise patterns.

For women, the study didn’t find meaningful differences. The researchers explained this may simply be because there weren’t enough women in the study to draw solid conclusions. Only two women were in the “laryngeal cancer with no other disorder” subgroup, which made reliable comparisons almost impossible.

Voice-based screening could help expand medical access, particularly for underserved populations who lack specialized voice clinics. Smartphone apps could analyze voice recordings, primary care doctors could conduct voice screenings during routine checkups, and telehealth platforms could flag concerning voice changes for further evaluation.

The technology would be especially valuable for high-risk groups: smokers, people with chronic acid reflux, those frequently exposed to vocal trauma, and individuals in voice-intensive professions. Regular voice monitoring could catch problems before they become serious health threats.

Recent artificial intelligence advances have made voice analysis increasingly sophisticated. Studies show that convolutional neural networks and deep learning models trained on spectrogram representations can classify laryngeal diseases, including early laryngeal cancer, with promising results using standard microphone recordings or even smartphone-captured voice samples.

The researchers emphasized that voice-based screening would supplement, not replace, traditional diagnostic methods. “While a definitive diagnosis still requires visualization, a validated AI-based voice screening tool could serve as a triage mechanism,” they noted.

Such a tool could identify individuals with subtle voice changes who may not otherwise seek care. Unlike the human ear, which may not reliably distinguish between subtle pathologic changes, an AI model can offer consistent and scalable voice analysis across diverse populations.

Limitations and What Needs to Come Next

Several limitations prevent immediate widespread use. The study’s small sample size (just 23 people with vocal fold lesions total) means these results need validation in much larger, more diverse populations before doctors can use them clinically.

The participant demographics also highlight why generalizability remains uncertain. Study participants were predominantly white (80.5%), heterosexual (79.3%), and female (63.2%). The authors stress that larger, more diverse populations are needed to validate the findings, which is a standard step in early research.

“The primary limitations of this study were the small sample size and participants’ incomplete lesion histories,” the researchers acknowledged. “Despite these limitations, the study provides valuable insights into the potential for voice biomarkers to serve as early indicators of vocal fold lesions.”

The most striking barrier for the selected features to be considered for a biomarker of vocal cord lesions is that, when researchers stratified data by sex, they found no statistically significant differences among women for either analysis group. “The power of these statistical tests was, of course, limited by the small sample sizes in some of our cohorts, most noticeably when comparing against the 2 cisgender women participants in the laryngeal cancer + no other vocal disorder cohort,” the team wrote.

Future research must recruit larger, more diverse study populations and explore additional voice features that might prove useful for women and underrepresented groups. Additionally, the sex of participants played a role in the results, which should be considered in future recruitment efforts to prevent biased datasets.

Still, the findings push scientists one step closer to making voice a routine health monitoring tool, as common as checking blood pressure or temperature. For millions of people at risk for vocal fold lesions, voice-based health screening could mean the difference between early detection and advanced disease.