Automatic recognition of inter-ictal epileptic activity in prolonged EEG recordings

Abstract

A method of automatic recognition and quantification of inter-ictal epileptic activity in the human EEG had previously been developed and tested using short recordings from awake subjects. This paper describes the adaptation of the method for use during overnight recordings in free-moving unattended patients, in combination with the already existing seizure monitoring system. EEGs were recorded from scalp and sphenoidal electrodes, using cable telemetry and a PDP-12 computer. The spike and sharp wave recognition method allowed the on-line analysis of 16 channels. A section of the 16-channel EEG including 1 sec before and 1 sec after each detected spike was saved on digital magnetic tape. Upon completion of the monitoring session, the tape was played back on the EEG machine, giving a discontinuous tracing of spike sections; this constituted a highly concentrated view of the inter-ictal epileptic activity, in traditional paper form. The spike sections were further analyzed by computer to determine and display on the computer terminal the spatial and temporal distributions of the epileptic activity, providing a complete synopsis of the recording. Several examples of the type of information available from this analysis are discussed in detail. False detection rates are given for 34 six hour recordings, indicating a high variability in the performance, mainly because of artefacts. It is concluded that the final computer displays could only be trusted after visual inspection of the EEG sections provided on paper. The variety of morphologies of artefacts appeared to preclude a total automatic elimination.