Artificial intelligence (AI) has made significant advancements in the field of healthcare, with one recent development being an AI system that can differentiate between diabetic and non-diabetic patients using voice recordings. Developed by Click Lab, this innovative technology can analyze and determine the differences in voice patterns between the two groups.
In a study published in the esteemed medical journal ‘Mayo Clinic Proceedings: Digital Health,’ Jaycee Kaufman’s research team from Canada’s Klick Labs revealed that their AI model can accurately identify type 2 diabetes using just 6 to 10 seconds of voice data, with an impressive accuracy rate of over 86%.
Type 2 diabetes is a chronic condition that occurs when the body’s ability to produce and use insulin is impaired. It accounts for approximately 90% of all diabetes cases and can lead to serious complications if left untreated, including cardiovascular and cerebrovascular issues.
To develop their AI model, the research team collected voice recordings from both healthy individuals and type 2 diabetes patients. Participants were asked to repeat fixed sentences several times a day for two weeks. By analyzing a total of 18,465 voice samples, the researchers identified 14 acoustic characteristics, such as breathing patterns, hoarseness, and roughness, that were distinct between the two groups. These acoustic features were then used to create an AI model capable of distinguishing between healthy and diabetic individuals based solely on voice recordings.
The results were highly promising, with the AI model achieving an accuracy rate of 89% for women and 86% for men. Remarkably, these accuracy rates were found to be comparable to traditional diabetes testing methods such as the fasting blood sugar test (85%), the glycated hemoglobin test (91%), and the oral glucose tolerance test (92%).
The significance of this AI voice diagnosis system lies in its potential to revolutionize the diagnosis of type 2 diabetes. Currently, the existing testing methods require patients to visit a hospital and undergo blood collection, which can be both time-consuming and costly. By utilizing voice recordings, the barriers to testing can be eliminated, enabling rapid and convenient diagnosis and treatment.
Lead researcher Kaufman emphasized the potential impact of this technology, stating, “Through signal processing, we were able to detect voice changes caused by type 2 diabetes, and these changes appear differently in men and women. This technology could change the way diabetes is tested.”
It is worth noting that diabetes often presents no noticeable symptoms in its early stages, leading to an alarming statistic where half of all adult diabetes patients worldwide (around 240 million individuals) are unaware of their condition. The development of this AI voice diagnosis system provides a promising solution to this issue, offering a non-invasive and accessible method of early detection.
Looking ahead, the research team aims to further validate their findings and expand the application of this technology to other areas, including the diagnosis of pre-diabetes, women’s health, and hypertension. As they continue their research, the potential for AI-driven healthcare solutions to transform the field of medicine becomes increasingly evident.
This breakthrough technology developed by Click Lab showcases the power of AI in healthcare and holds immense promise for improving the diagnosis and management of type 2 diabetes. With its accuracy and convenience, this AI voice diagnosis system has the potential to enhance healthcare accessibility and ultimately, improve patient outcomes.
Click Lab is developing an AI that can differentiate between diabetic and non-diabetic patients using 6 to 10 seconds of voice.
Kaufman: “AI voice diagnosis system will break down barriers to diagnosing type 2 diabetes”
Comparing the accuracy of AI diabetes diagnosis using voice and existing testing methods. (Photo = Click Labs)
[서울파이낸스 이도경 기자] Artificial intelligence (AI) has been developed to investigate, analyze, and determine the differences in voice between non-diabetics and type 2 diabetic patients using voices recorded on smartphones.
In an article published in the medical journal ‘Mayo Clinic Proceedings: Digital Health’ on the 19th, Jaycee Kaufman’s research team from Canada’s Klick Labs said that this AI model can record 6 to 10 seconds of video. It is said that it able to distinguish type 2 diabetes using voice with an accuracy of over 86%.
Type 2 diabetes is a chronic disease that occurs when the body’s ability to secrete or use insulin is impaired and accounts for approximately 90% of all diabetes patients. Type 2 diabetes, if left untreated, can develop into cardiovascular and cerebrovascular complications, which require proper management and treatment.
The research team recorded fixed sentences (Hello, Are you?) at least six times a day for two weeks.
As a result of analyzing 18,465 recorded samples, the research team discovered 14 acoustic characteristics such as breathing, hoarseness, and roughness between healthy people and type 2 diabetes patients, and extracted and used these to identify type 2 diabetes by voice. created an AI model to distinguish between diabetes.
The research team said this AI model was able to determine type 2 diabetes with an accuracy of 89% for women and 86% for men. The research team explained that this was comparable to the accuracy of the fasting blood sugar test (FBG, 85%), the glycated hemoglobin test (A1C, 91%), and the oral glucose tolerance test (OGTT, 92%).
The research team evaluated that the development of this system will bring significant changes to the diagnosis of type 2 diabetes.
According to the International Diabetes Federation (IDF), diabetes has no clear symptoms in the early stages, so one in two adult patients worldwide (around 240 million people) are unaware that they have the a disease.
Despite the need for rapid diagnosis and treatment, diabetes diagnosis such as FBG, A1C, and OGTT all required a visit to a hospital to collect blood, which was time consuming and costly, creating a barrier to testing.
“Through signal processing, we were able to detect voice changes caused by type 2 diabetes, and these changes appear differently in men and women,” said researcher Kaufman. “This technology could change the way diabetes is tested.” he said.
“This test method has the potential to remove the barriers to current time- and money-consuming testing methods,” he said. “In the future, we will conduct research to further verify the results of this study and expand negative diagnosis to pre-diabetes, women’s health, and hypertension.” “I will,” he added.
Copyright © Seoul Finance Reproduction and redistribution prohibited.
#diagnoses #diabetes #voice #emerges.. #Accuracy
