Catholic Medical Center Develops Voice-Based Artificial Intelligence for Dementia Diagnosis
Sep 10, 2024
A research team led by Professor Ko Tae-hoon (corresponding author, medical informatics class at Catholic University Medical School) and Professor Kim Dong-min (first author, research professor at Catholic University Medical School's medical informatics class) at the Catholic University of Korea have developed a voice-based dementia classification artificial intelligence that can respond to both Korean and English.
In the case of developing voice-based artificial intelligence for classifying the degree of cognitive impairment of patients, most of them develop artificial intelligence for each type of language. However, there are limitations in collecting AI learning datasets (structured collections of data organized and stored together for analysis or processing) by language type, and there have been difficulties in developing a number of individual AIs.
The research team succeeded in developing artificial intelligence to classify mild cognitive impairment groups and dementia groups at a high level in both languages using Korean and English voice datasets. In this study, the optimal artificial intelligence was developed by comparing various deep learning models using Mel Sepctrogram, which converted the Korean/English speech of the mild cognitive impairment group and the dementia group into a frequency form.
Professor Ko Tae-hoon said, "The artificial intelligence developed this time can be the basis of a voice biomarker-based dementia diagnosis solution that can be applied universally regardless of each language type in the future." he said.
Meanwhile, the research was conducted with the support of the Artificial Intelligence Brain Science Project Group of the Catholic Central Medical Center's Basic Medicine Project Promotion Team, and was recently published in the prestigious medical information journal 『Computers in Biology and Medicine』.
In the case of developing voice-based artificial intelligence for classifying the degree of cognitive impairment of patients, most of them develop artificial intelligence for each type of language. However, there are limitations in collecting AI learning datasets (structured collections of data organized and stored together for analysis or processing) by language type, and there have been difficulties in developing a number of individual AIs.
The research team succeeded in developing artificial intelligence to classify mild cognitive impairment groups and dementia groups at a high level in both languages using Korean and English voice datasets. In this study, the optimal artificial intelligence was developed by comparing various deep learning models using Mel Sepctrogram, which converted the Korean/English speech of the mild cognitive impairment group and the dementia group into a frequency form.
Professor Ko Tae-hoon said, "The artificial intelligence developed this time can be the basis of a voice biomarker-based dementia diagnosis solution that can be applied universally regardless of each language type in the future." he said.
Meanwhile, the research was conducted with the support of the Artificial Intelligence Brain Science Project Group of the Catholic Central Medical Center's Basic Medicine Project Promotion Team, and was recently published in the prestigious medical information journal 『Computers in Biology and Medicine』.
|
bellho@sportschosun.com
