Abstract Details

Title Determining In-Depth Information of Mild Cognitive Impairment: Clustering Using Unsupervised Machine Learning

Topic Aging, Dementia, and Behavioral Neurology

Presentation(s) Aging and Dementia Posters (7:00 AM-5:00 PM)

Poster/Presentation
Number 067

Objective To determine clusters of aging Thais with early mild cognitive impairment (MCI) using unsupervised machine learning based on Montreal Cognitive Assessment (MoCA)

Background MCI distinguishes four clinical subtypes: amnestic MCI-single domain, amnestic MCI-multiple domains, nonamnestic MCI-single domain, and nonamnestic MCI-multiple domains. However, in-depth information of multiple domains is lacking and the controversy of possible undiscovered subtypes and the appropriate cut-off point is on-going. Clustering is an unsupervised machine learning technique that can reveal subgroups within heterogeneous data that might be difficult to recognize.

Design/Methods Our prospective cohort enrolled healthy Thai adults aged ≥60 years. Cognitive performance was evaluated by the validated Thai version of MoCA. Those who scored between 23 and 27 (cut-off point ± 2) were included in the analysis. Normalized total score of each cognitive domain served as inputs for K-mean cluster algorithm. In addition, data was split into two and clustered independently to examine the consistency of the model.

Results Among 2799 participants enrolled, 1592 (56.9%) scored between 23 and 27 and were included in the analysis. Models showed consistency and 6 clusters were found: 20% had pure memory impairment, 29% had congruous language-memory impairment, 19% had predominant language with memory impairment, 15% had congruous abstraction-memory impairment, 11% had congruous abstraction-language impairment, and 6% had predominant abstraction with language-memory impairment. There were no significant differences in mean MoCA score between groups. Congruous abstraction-language and predominant abstraction with language-memory impairment were twice more likely to occur in female than male and up to 2.5 times more likely to occur among those with highest education of high school or less than their counterparts (p<0.001).

Conclusions Six clusters surrounding language, abstraction, and memory impairment, mostly with multiple domains involvement, were demonstrated among aging Thais with early MCI. A longitudinal study is ongoing to identify differences in clinical significances and prognosis between each cluster.

Authors/Disclosures
Akarin Hiransuthikul, MD (Faculty of Medicine, Chulalongkorn University) PRESENTER	Dr. Hiransuthikul has nothing to disclose.
	No disclosure on file
Achitpol Thongkam	Mr. Thongkam has nothing to disclose.
	No disclosure on file
	No disclosure on file
	No disclosure on file
	No disclosure on file
	No disclosure on file

�鶹��ýӳ��

�鶹��ýӳ��

Abstract Details