Authors : Sunghyoun Cho, Seonghun Yu, Heekyung Jin and Haewon Byeon
Abstract: This study identified the visuospatial and verbal characteristics of Mild Cognitive Impairment (MCI), developed a MCI prediction model based on the random forest technique with using the identified characteristics and provided basic data for diagnosing and intervening MCI. The subjects of study were 245 elderly over the age of 65 (174 healthy elderly and 71 elderly with MCI). Visuospatial function was assessed by using Rey-Osterrieth Complex Figure (RCFT-copy) and linguistic ability was assessed by using Korean Boston Naming test (K-BNT) and Korean-Mini Mental State Examination (K-MMSE). The prediction model was developed by using the random forest technique and its predictive power was compared with models developed by logistic regression analysis and decision tree. As the result of independent t-test, healthy elderly group had significantly higher mean value of RCFT-copy and K-BNT than MCI group. There was no significant difference, however in reading and writing ability. The prediction model developed by using the random forest technique showed that major predictors of it were RCFT-copy and K-BNT and the accuracy of it was 66.1%. Random forests had better predictive power than logistic regression analysis and decision tree. In this study, visuospatial ability and confrontation naming ability were significant predictors of MCI. In order to investigate the causal relationship between visuospatial functions and confrontational naming ability and MCI, longitudinal studies are required to be conducted in the future.
Sunghyoun Cho, Seonghun Yu, Heekyung Jin and Haewon Byeon, 2018. The Development of Mild Cognitive Impairment Prediction Model based on Random Forest with Focusing on Visuospatial Function and Language Ability. Journal of Engineering and Applied Sciences, 13: 2885-2888.