Objective: To compare the performance of an EEG data selection/reduction algorithm for epileptic EEGs on ambulatory and non-ambulatory recorded data to confirm that acceptable performance is achievable in ambulatory recordings despite the presence of overt artifacts.
Methods: A total of 167 hours of EEG data containing 899 marked interictal events is analysed to determine the percentage of events correctly recorded (the sensitivity) and the amount of data reduction achieved.
Results: A better sensitivity-data reduction trade-off is found in the ambulatory recorded data. This may be unexpected as ambulatory recordings are known to contain large numbers of artifacts, but is accounted for by these artifacts being easily detected and discarded, improving the data reduction.
Conclusions: Satisfactory performance levels are found in both data types, no degradation is present with ambulatory recordings.
Significance: Demonstrates that the processing of EEG data for wearable EEG applications is feasible without a loss in performance compared to traditional inpatient EEG usage.