Tagging EEG features within exam reports to quickly generate databases for research purposes.

Objective: assess the effectiveness of a new method for classifying EEG recording features through the use of tags within reports. We present feature prevalence in a sample of patients with toxic-metabolic encephalopathy and discuss the advantages of this approach over existing classification systems.

Methods: during EEG report creation, tags reflecting background activity, epileptiform features and periodic discharges were selected according to the findings of each recording. Reports including the tags have been collected and processed by the EEG report parser script written in PHP language. The resulting spreadsheet was analysed to calculate the prevalence and type of EEG features in a sample group of patients with toxic-metabolic encephalopathy.

Results: tag checking and extraction were very little time-consuming processes. Considering 5784 EEG recordings performed either in inpatients or outpatients over 2 years, toxic-metabolic aetiology was tagged in 218 (3.8 %). The most frequent background feature was severe slowing (5-6 Hz frequency), occurring in 79 (36.2 %). Epileptiform abnormalities were rare, reaching a maximum of 10 (4.6 %). Triphasic waves were tagged in 43 (19.7 %) recordings.

Conclusions: tagging and parsing processes are very fast and integrated into the daily routine. Sample analysis in patients with toxic-metabolic encephalopathies showed EEG slowing as the prevalent feature, while triphasic waves occurred in a minority of recordings. Existing software such as "SCORE" (Holberg EEG) requires the replacement of the currently used software for EEG reporting, minimizing additional costs and training. EEG Report Parser is free and open-source software, so it can be freely adopted, modified and redistributed, allowing further improvement and adaptability.