AI can hear early signs of cancer in your voice, study finds

A study published in the journal Frontiers in Digital Health, found that artificial intelligence (AI) retains the potential to detect abnormal growths on the vocal cords, from benign nodules to early stage detection of laryngeal cancer by examining short voice recordings.

These findings could further support finding an easier method to diagnose cancerous lesions on the vocal records, known as folds.

The study’s lead author in clinical informatics at Oregon Health and Science University in the United States, Phillip Jenkins, said, “With this dataset we could use vocal biomarkers to distinguish voices from those without such lesions.”

Early detection of throat cancer is crucial for better outcomes and survival rates

Larynx cancer can affect more than a million people worldwide and kill approximately 1,00,000 each year. It is the 20th most common cancer in this world.

The key factors of this disease are smoking, alcoholic usage, and certain strains of HPV (human papillomavirus) and survival rates vary from 35 to 90 percent. It particularly depends upon the early diagnosis of the disease.

The most common warning signs for laryngeal cancer are hoarseness and changes in voice that last more than three weeks. More symptoms include sore throat, swelling, lump in the neck, ear and throat pain.

Meanwhile, early detection is crucial to improve survival rates through timely treatment.

Future steps for AI driven diagnosis

The research team has examined about 12,500 voice messages from 306 people across North America. Some of the subtle acoustic patterns were found, such as changes in pitch and loudness clarity.

The teams have found no comprehensible difference for men in the acoustic analysis and pitch between those with healthy voices and cancer.

While no significant results were found due to the limitations in its dataset.

Senior author Phillip Jenkins was of the view that the next step is to train AI models on larger dataset and test them in clinical settings to analyze whether the system provides results that work well for both men and women.

He stated, “Based on our findings, I estimate that with larger datasets and clinical validation, similar tools to detect vocal lesions might enter pilot testing in the next couple of years.”