The overall purpose of the Phase I project was to develop and refine a testing method to assess the work readiness of individuals who work in occupations that require alertness and acute judgment for the operation of high-risk technology. Our approach to this issue was to measure neurophysiological and performance variables recorded from subjects performing attention-demanding cognitive tasks, and to apply neural network pattern recognition analysis to detect subtle multivariate differences between alert and impaired states. During Phase I we sought to test the adequacy of the key signal processing and pattern recognition methods that the device would use, refine a test battery, and define the necessary functionality to design a device suitable for use in work environments. The work completed in Phase I was more than that specified in the original aims. Specifically, we analyzed a larger and more recent database that had more (N=9) subjects and that had alcohol as well as fatigue stressors, rather than the older one described in the proposal that had fewer (N=4) subjects and only fatigue as a stressor. After completing descriptive statistics and individual subject pattern recognition studies, we focused on testing the feasibility of the key underlying principle of the Work Readiness Neurometer; namely, that it is possible to distinguish between alert and impaired states in individuals for whom there is no prior sample of the impaired state, and on other key ideas such as use of task-related EEG rather than resting EEG, and of differentiating different sources of impairment. Due to the limited scope of a Phase I project, only a small group of subjects were analyzed. Thus, these results must be treated as preliminary and limited in their generalizability. Nonetheless, they are remarkable in that they suggest that impairment-related changes in the EEG are highly reproducible and similar across subjects. They are the first demonstration that a generic (group) EEG pattern recognition network can successfully be applied to a new subject to determine whether he or she is in an Intoxicated or Fatigued state. These results have served to clarify several important issues with respect to the feasibility of developing a work readiness test based on neurophysiological measurements. First, we found that task related EEG signals were more sensitive to states of impairment than were eyes-open resting EEG signals, indicating that the test should require that subjects perform some attention demanding task. Second, we found that both EEG signals and test performance measures during a simple working memory task were more sensitive to states of impairment than were either behavioral measures or EEG signals during a perceptuomotor tracking task, suggesting that a simple test based on the WM task might alone be adequate for work readiness testing. Third, we found that EEG features provided sensitive indicators of fatigue even under conditions where task performance variables were unaffected. Fourth, we found that the EEG signature of impairment from alcohol was dramatically different from that associated with Fatigue. Fifth, we found that we could dissociate theta band EEG signals related to task difficulty from those related to fatigue using topographic criteria. Finally, we found that it was possible to classify states of impairment in subjects without prior knowledge of what their individual EEG signals look like during impaired states. These findings establish the basic feasibility of creating a sensitive neurophysiological measure of work readiness that would be suitable for use in conventional work environments.
One Rincon Center, 101 Spear Street, Suite 203, San Francisco, CA 94105