PolyU develops an AI system for assessing the risk of dementia
Dementia is a disease of decline in brain functions that will affect the patient’s memory, language, judgement as well as daily life. Early identification of people with a high dementia risk is important for timely intervention and support. A team from PolyU’s School of Nursing has developed an artificial intelligence (AI) system that can assess the risk of dementia by using health data of the elderly.
Currently one of the most commonly used cognitive assessment tools is the Mini-Mental State Examination (MMSE), which is a questionnaire that is administered when symptoms of memory decline have occurred. However, according to Professor Thomas Choi of the School of Nursing who led the research, early administration at the asymptomatic stage and repeated use of the same questionnaire would lead to a “practice effect” that degrades the effectiveness of the MMSE when it is used in later stages.
PolyU’s AI assessment system uses the health data of the elderly, such as basic information and a health index like age, gender, blood pressure, teeth condition and nutritional assessment to screen the risk of dementia. The team collected data from over 2,000 elderly aged 77 on average from 2008 to 2018 for assessment, and the accuracy of screening dementia reached 88%.
Dr Rick Kwan, Assistant Professor of the School of Nursing, said the data needed for assessment is already known by community centres or residential care homes, therefore they may acquire the assessment results by simply uploading the data to the system.
The research team has secured about HK$3 million from the Innovation and Technology Fund, and has applied for a patent in the US. Prof Choi said they will apply to join the Public Sector Trial Scheme under the Innovation and Technology Commission to develop smart health information systems for elderly service centres.
A media interview was held recently to introduce this AI system. Have a look at the media reports to learn more about the research project:
