February 23, 2023

bizwatch

By Mi-Ran,Kwon rani19@bizwatch.co.kr

 

Dementia patient detection accuracy at 87%… Outperforming the 70% average
The AI technology will be presented at the ICASSP conference this coming June

 

 

An AI technology for Alzheimer’s detection, independently developed through a collaboration between a research team at Konkuk University Hospital and a domestic AI startup, has taken first place in a global Alzheimer’s disease AI detection competition.

 

On the 23rd, the joint AI research team—led by Professor Shin Jeong-eun (Department of Otorhinolaryngology-Head and Neck Surgery, Konkuk University Hospital) and Professor Kim Eun-yi (Department of Computer Engineering, Konkuk University), alongside team members Kim Ryong-bin, Kim Hyun-seo, Oh Ye-rim, Jeon Hyo-jin, and Jung Hyun-taek—along with AI startup Voinosis, announced their victory at The MADRess Challenge. The competition was held at the 2023 International Conference on Acoustics, Speech, and Signal Processing (ICASSP).

 

Now in its third year, The MADRess Challenge is an international competition where participating teams evaluate and compare the performance of their independently developed AI models for Alzheimer’s disease. Teams are required to determine the presence of cognitive impairment within approximately one minute by analyzing the voices of patients with dementia and mild cognitive impairment. They must also predict the patients’ scores on the MMSE (Mini-Mental Status Exam), a standard cognitive screening test.

 

Notably, to assess the technology’s potential for global application regardless of spoken language, this year’s competition evaluated the AI’s multilingual performance (English/Greek) by relying exclusively on the acoustic characteristics of the voice.

 

While most participating teams averaged a 70% accuracy rate for detecting dementia patients and an error margin of around 5 points for dementia severity, the AI developed by Konkuk University Hospital and Voinosis recorded an impressive 87% detection accuracy and an error margin of just 3.7 points for predicting severity. Their AI secured the win by demonstrating significantly superior accuracy.

 

As a result, the Konkuk University Hospital research team has been invited to the ICASSP conference—the most prestigious academic event in the signal processing field—to be held in Greece this June, where they will present their AI technology to world-renowned scholars.

 

Professor Shin Jeong-eun expressed her gratitude, saying, “I am delighted that the hard work of all our team members has yielded such wonderful results. I hope the AI solution we have developed will greatly contribute to creating a dementia-free world by allowing us to identify patients with early-stage cognitive impairment more quickly and easily.”

 

 

original article : Konkuk University Hospital and Voinosis Win Global AI Competition for Alzheimer’s Detection