TECHNIQUES TO IDENTIFY CLINICAL CONTEXTS DURING AUTOMATED DATA-ANALYSIS

The interpretation of automatically collected data to produce intellig ent alarms and identify particular conditions is nearly impossible wit hout identifying the specific context in which the data are obtained. Shifts in clinical context occur because of changes in the patient's p hysiologic state, or due to the passage of time, or due to changes imp osed by therapeutic intervention such as surgery. Techniques to identi fy such changes in clinical context are discussed with particular atte ntion to the application of cluster analysis, discriminant analysis, a nd statistical predictors. An example of these analyses applied to EEG data is presented, showing an unexpected hysteresis of EEG behavior i n response to an hypoxic challenge.