Researchers have shown that vocal cord lesions can be discovered through the patient’s voice, and these results open the door to a new application of artificial intelligence to learn the early warnings of throat cancer through sound recordings, where such lesions can be benign, and it may also represent the early stages of throat cancer.
The study was conducted by researchers from the Department of Information, Clinical Epidemiology, Oregon University of Health and Science in the United States, and its results were published in the Journal of Digital Health Prospects (Frontiers in Digital Health) on August 12, and was written by the Yorik Alrt.
Throat cancer constitutes a large general healthy burden, and the number of cases of throat cancer around the world was estimated in 2021 at about 1.1 million cases, and he died due to nearly 100,000 people.
The risk factors include this cancer, smoking, wine, and infection with HPV. The survival rate for throat cancer ranges between 35% and 78% over 5 years after treatment, depending on the tumor stage and its location inside the throat.
Early detection of cancer is very important for the patient’s safety. Throat cancer is currently diagnosed through the endoscopy of the nose with video and dazzling, which are arduous procedures and require surgical intervention. It may take time to reach a specialist who is able to conduct these tests at a time, which causes a delay in diagnosis.
“We show here that using a set of data, we can use vocal vital indicators to distinguish the sounds of patients with vocal cord lesions from those who are not infected.”
Voice messages
Jenkins and his colleagues shared members of the “Bridge to AI Intelligence Intelligence” project (Bridge2ai-Voice) in the “Bridge” Bridge “Bridge” Bridge “Bridge2ai” of the American National Institute of Health, an endeavor to apply artificial intelligence to complex biomedical challenges.
They analyzed the differences in the tone, the class, the size, and the clarity within the first version of a general data collection of artificial intelligence related to sound, with more than 12 thousand audio recording of 306 participants from all over North America.
A few patients had well -known throat cancer, benign deformities in the vocal cords, or two other cases in the throat: spastic speech defect and monochromatic vocal cord paralysis.
The next researchers ’step is to use new algorithms on more data and test them in hospitals on patients’ sounds.
“To move from this study to an artificial intelligence tool that recognizes the lesions of the sound cords, we will train models using a larger data collection of audio recordings, classified by specialists, then we need to test the system to ensure that it works well on an equal basis for women and men,” says Jenkins.
Experiments are currently on sound -based health tools, and based on the results of studies, Jenkins expects that with the availability of larger data collections and clinical achievement, similar tools may enter to detect the lesions of the audio folds, the experimental test stage during the next two years.