EEG-based data mining approach for measuring cognitive information processing performance

Abstract

Human cognitive information processing is essential in daily life. It consists several fundamental cognitive processes such as attention, memory, and decision / execution. Neuropsychological test is the most wildly-used measure for human cognitive information processing performance (CIPP). However, neuropsychological test in case of measuring the cognitive speed can be affected by irrelevant factors such as physical performance. Furthermore, neuropsychological test shows interpersonal variation so that it may not an objective measure to quantify the effect of the environment to human CIPP. Therefore, we proposed data mining approach to develop EEG-based measure which more directly quantifies the cortical activity than final end result of neuropsychological test and shows less interpersonal variation for objective measurement of environmental effect. To compare the accuracy of EEG-based measure and that of neuropsychological test, thermal condition was adapted as surrogate conditions, which correspond to CIPP level. 10 healthy male subjects repeated three neuropsychological tests with EEG measurement in three thermal conditions of $24^circ C$, $3^circ C$, and $36^circ C$, which correspond to three CIPP levels of 'Good', 'Bad', and 'Worse', respectively. The three neuropsychological tests were contingent continuous performance task (attention), visual pattern span (memory), and Wisconsin card sorting test (decision / execution). As a result of training through C4.5 algorithm, central (C4) $\beta$ during WCST and mid-frontal (Fz) $\gamma$ during VPS were selected to classify three thermal conditions. The accuracy of the proposed EEG-based measure was 83.3% and outperformed that of neuropsychological test which was 43.3% through leave-one-out cross-validation. The proposed EEG-based measure can be utilized to develop CIPP-promoting environment.