The electroencephalography (EEG) signal has been used for many decades now to diagnose cerebral dysfunction, mainly through visual analysis of the recordings. There have been several attempts to make an automatic analysis of the EEG signal by creating different algorithms for signal analysis. Some of the more well-defined graphoelements of the EEG such as spikes and sharp waves and also rhythmicity of ictal episodes can be detected with different computer applications, but the major part of the EEG will be subjected to a visual analysis by trained experts. This means that the interobserver agreement in EEG interpretation is only moderate as recognized by the authors. There are several reasons for this such as ability, training, and experience of the EEG reader as well as the free-text format of the EEG descriptions.
The authors of the report formed a working group of EEG experts from 15 European countries, and the aim was to construct software for characterizing EEG and ictal clinical events. The EEG reader should choose from predefined terms and at the same time generate a report and fill information into a database. This would in the end increase interobserver agreement, construct a multinational database for further research projects, and be a useful tool in education and training of EEG readers.
So have they succeeded with the software SCORE? When you look into the result and start with the main elements put in the flowchart of SCORE, you can conclude that the list is complete. It ranges from patient's personal data, referral data, recording conditions, modulators/procedures, background activity, sleep and drowsiness, interictal findings, episodes, physiologic patterns, patterns of uncertain significance, EEG artifacts, polygraphic channels, and finally diagnostic significance. Even the referral data can be classified into different reasons for referral, which I find most important with the possibility to pool data in multicenter studies.
The terminology within the different main elements, for example, interictal graphoelements, names of episodes, and ictal EEG patterns is well chosen and should not be misunderstood by anyone with basic knowledge in EEG interpretation.
Under the heading of episodes you will find a list of clinical episodes as well as electrographic seizures. Following the clinical practice in most EEG laboratories, SCORE makes it possible to group and describe several clinical episodes (seizures) under the same heading, if the physician considers them as manifestation of the same phenomenon. The ictal semiology is well covered in the list of clinical signs during episodes provided in SCORE, so also is the list of different ictal EEG patterns.
SCORE also takes into consideration the special aspects of neonatal recordings. A special neonatal matrix is loaded by the software instead of the “background activity” and “sleep.” The main elements in the neonatal matrix consist of behavioral stages and temporal and spatial organization. This matrix makes it possible to use SCORE in the analysis of the continuous EEG monitoring with different on-line trend algorithms in the neonatal wards.
Following the scoring of the recording a report is automatically generated. The report has a flexible format, which means that it can be easily edited and free text can be added. In order to use the full potential of the reporting module you would need to create an interface with the local digital patient chart.
Preceding the report module there is a part of SCORE that I am especially fond of and that is the part where the physician will put the scored EEG features into the clinical context. Sometimes you will find an EEG report consisting mainly of a detailed description of the different graphoelements of the EEG signal, but at the end you do not find any conclusion or actual interpretation of the EEG setting the data in the clinical context. In SCORE you cannot pass this issue and it enhances the indication for the EEG recording. What is the clinical question—why was the EEG recording performed?
A free shareware version of SCORE was provided by Holberg EEG AS company to our EEG laboratory, Neuroscience Center, Academic Hospital, Uppsala, Sweden, for testing, so I have personal experience of the software and it has also been tested by several of my younger colleagues and also by residents in training in clinical neurophysiology at our department. One initial apprehension was that the use of SCORE would significantly increase the time of reading an EEG. It turned out that this was not the case. The normal EEG was just as easily read and the software allows short cuts without missing the point of scoring through the main parts. The implementation of the software into the routine reading of EEG was also easily done. We did not have an interface to the local digital patient chart, thus making it somewhat difficult to use the full potential of the reporting module. It is also clear that the software has to be translated into the native language of the EEG laboratory in order to be used in the routine—not at least for medicolegal reasons.
Apart from these minor adjustments the software works just as it is meant to work, and especially in the situation of introducing new colleagues into the reading of EEG recordings I think it will be a great instrument to use for that purpose. The authors mention in the discussion in the article that in addition to the definition, typical examples of EEG samples (screen shots) will be accessible directly in the software, which of course adds the great potential for SCORE in training neurophysiologists.
I agree with the authors that the use of SCORE will increase the interobserver agreement, and the software will offer a possibility for quality control and audit between EEG laboratories.
In the current version of the software the database is produced locally. There are plans to make an international database, where centers wishing to participate can upload their data. The legal background for data transfer has to be clarified before proceeding to this. This will provide an opportunity for launching different projects concerning the diagnostic value of EEG, but the future will have to show whether this will be a realistic enterprise—still the local database as such can of course be useful for different projects within the local EEG laboratory.
My final impression of the software SCORE—having read the present article as well as tested the software provided by Holberg EEG AS version 18.104.22.16807—is positive, and I think that many EEG laboratories in Europe will start using the software once the translation into the native language has been accomplished.