Online Program

Return to main conference page
Tuesday, January 7
Tue, Jan 7, 9:00 AM - 10:45 AM
West Coast Ballroom
Statistical Learning Methods for Health Care Innovation

A dynamic optimization screening system for mild cognitive impairment based on machine learning model (307818)

*Guohong LI, Shanghai JiaoTong University School of Medicine 

Keywords: Machine learning, Mild cognitive impairment, Screen, Computerized neuropsychological assessment devices

[Objectives] Dementia is one of the most severe diseases which may threat the quality of old age. Mild cognitive impairment (MCI) is a pre-clinical stage between natural ageing and dementia. Those who has been diagnosed MCI may have high risk to getting to Alzheimer’s disease. Early screening and interventions for MCI patients might improve or prevent their cognitive impairment to some extent, and also delay the course of dementia. This study aims to design a self-evaluable computerized cognitive assessment system for screening MCI in a large range, evaluate its reliability and validity, and compare the screen accuracy among different machine learning models. [Methods] This study based on the cognitive design system, and referring to the most common used paper-and-pencil cognitive tests, designed a synthetic self-administrate computerized cognitive assessment system which contains scenario. A group of over-55-years-old samples which come from Shanghai and Henan, were selected to assess their cognitive status by well-trained investigators. Using intrinsic consistency analysis and factor analysis to assess reliability and validity respectively. Four machine learning classification algorithms were compared by their accuracy and area under the curve(AUC). [Results] 359 participants were included, the mean age of those participants was 63 years old, and 82.70% of them were less than secondary graduation. The intrinsic consistency and KMO of this system were 0.84 and 0.78 respectively, and Bartlett’s sphericity test was significant(P<0.05), thirteen common factors could explain 75.10% of the system. Among four machining learning models, the best classification model was naïve Bayesian classification model, and its accuracy was 88.05%, AUC was 0.941. [Conclusions] The new designed computerized cognitive assessment system has been proved to be reliable and valid. It’s suitable for early cognitive impairment screening in communities, which can assist the general practitioner, improving health resources utilization efficiency.