서울대학교 생물정보학 협동과정 주최로 세미나를 아래와 같이 열고자 하오니 많은 참여 바랍니다.

세미나는 Zoom을 통한 온라인 강의로 예정되어 있으며 강의에 참여하고자 하시는 분께서는 아래의 링크로 참여 부탁드리겠습니다.

*줌으로 참석하기 위해서는 "참석자로 회의 참가" --> 이름, 성, 이메일 주소, 이메일 주소 확인을 입력하시고 등록하시면 회의 참석이 가능합니다.

일시: 2023년 3월 9일 목요일 오전 11시

연사: 김상철 박사님 (국립보건원)

Zoom link:　http://snu-ac-kr.zoom.us/s/94919995246

Prediction of Progress to Dementia using Subjective Memory Impairment and Dementia Characteristics with Brain MRI, Cognition Test, APOE4 Genotype.

Hye Ryeong Nam¹, Sang Cheol Kim¹*

¹Division of Healthcare and Artificial Intelligence, National Institute of Health, Korea Disease Control and Prevention Agency, Cheong-Ju, Republic of Korea.

Abstract

Brain atrophy and cognitive function decline are known to precede dementia. In addition, the apolipoprotein-e4 allele(APOE e4) carrier also increased the risk of dementia.

Our goal is to use clinical data, brain atrophy, cognitive function, the APOE e4 genotype, and other factors to develop a prediction model for dementia progression. Additionally, utilizing the two years of follow-up data, it is intended to evaluate the predicted value and the altered actual condition of cognitive function.

We used participants in the Alzheimer's disease study (BICWALZS) who had chronic cerebrovascular illness to test biobank innovations. We categorized 686 participants into three groups: 78 with subjective memory impairment (SMI), 392 with mild cognitive impairment (MCI), and 207 with dementia. These subjects were free of vascular dementia and other disorders that affect cognition function. Each individual had a 3T brain MRI, performed a cognitive test, and had their APOE genotypes determined. Feature selection analysis used the stepwise correlation based on the forward method. Our model can prediction to cognitive status using the K-Nearest Neighbor (KNN) and Support Vector Machine(SVM) method. In addition, we performed the Shapley additive explanations(SHAP) method to study each feature's importance and to understand prediction models.

The stepwise method selected features of delayed recall memory in cognitive function, gray matter volume of the temporal lobe, and APOE e4 carrier. With the KNN approach and the SVM method, the prediction model's accuracy was 0.99 and 0.98, respectively. Our validation test included 392 MCI individuals with 2-year follow-up information. The reactive change value was the same as the forecast value for the four participants who had progressed to dementia. The three most significant factors influencing the altered cognitive state discovered utilizing the SHAP technique were the gray matter volume, blood pressure, and APOE e4.