Effects of Lorazepam on the Automatic Online Evaluation of Sleep EEG Data in Healthy Volunteers

In earlier publications we described an automatic algorithm to detect rapid eye movement (REM) sleep from a single-channel EEG recording without using EMG or EOG information. This system consisted of an artificial neural network operating on the basis of preprocessed EEG data and was composed to provide a maximum of robustness for online applications. In the present study the influence of acute administration of lorazepam on the performance of the REM detection procedure was evaluated. Following an adaptation to laboratory conditions, sleep EEG data were obtained from healthy subjects in three nights each. On the evening of the second night the volunteers received a single dosage of 2.5 mg Lorazepam; the other two nights were drug-free. The sleep profile and the quantitative EEG data reflected the known changes following acute administration of benzodiazepines: during the treatment night the amount of non-REM sleep and the relative power of the EEG signal in the beta and gamma frequency bands was increased relative to the first night, while the amount of REM sleep was reduced. The night of drug discontinuation still showed some characteristics of the treatment night. The discordance rate of the REM detection algorithm relative to the manual evaluation ranged from 9 % to 14.2 % for the different nights. Surprisingly, the percentage of correctly classified time periods was even higher for the lorazepam night as compared to the other nights.