For our investigation we used EEG recordings obtained from a sedated intensive care patient that was included in the clinical trial “Validation of methods for monitoring nociception and pain prediction in the ICU”. This study was performed in accordance with the ethical standards as laid down in the 1964 Declaration of Helsinki and its later amendments. It was approved and the requirement for written informed consent was waived by the Institutional Research Ethics Committee of Charité - Universitätsmedizin Berlin (vote number: EA1/151/16). The trial was registered prior to patient enrollment at the German Clinical Trial Register (DRKS00011206; Principal investigator: Dr. Falk von Dincklage, Date of registration: 11.01.2017).
SEDLine EEG recording
When using the SEDLine, an electrode sensor is attached to the patient’s forehead. The sensor consists of six electrodes that record brain electrical activity from recording positions Fp1, Fp2, F7 and F8 according to the international 10–20 system with the reference and ground electrode placed on the center of the forehead (around Fz). The SEDLine monitor processes the EEG information to calculate the PSI. It also displays the changes in spectral EEG power over time as a density spectral array (DSA) on the monitor’s display to provide the anesthesiologist with additional information regarding the oscillatory composition of the EEG. Further, the raw EEG can also be displayed on the monitor. In order to optimally visualize the EEG trace, the anesthesiologist can adjust the time-scale of the EEG as well as the amplitude resolution. The scaling is either 15 mm or 30 mm per second and the amplitudes can be adjusted to 1, 2, 3, 5, 10, 25, 50, or 100 µV per mm.
SEDLine EEG export
It is possible to export a recorded EEG file from the SEDLine Root device in European Data Format (EDF file, .edf extension).Using the export function, the EEG can be exported to a USB-drive as an EDF file.
The EDF files consist of a header record and a data record. The data record contains the time series data, i.e., the amplitude values recorded at equally-spaced time points. In exported SEDLine EEG files the data record contains numeric information from the four frontal EEG channels. The header record contains information regarding electrode positions, patient and time information etc. The sample rate is not given explicitly, but can be obtained by dividing the value of the header variable ‘samples’ by the value of the header variable ‘duration’. Hence, one sample rate value can be derived that reflects the sample rate for the entire recording. The sample rate is defined at the beginning of the recording.
Change of EEG display protocol
Clinical EEG recording
We suspected that the choice of parameter settings on the SEDLine display during EEG recording may influence the raw EEG that can be exported to the EDF file. Thus, we investigated this possible effect using EEG recordings obtained from a sedated intensive care patient. The patient used for EEG measurement presented the following characteristic: male, 55 years of age, light analgosedation using propofol, clonidine as well as methyl-lorazepam in the post-operative setting. Because we were only interested in the question of how the monitor settings influence the exported EEG, one patient was considered sufficient. We decided on a patient under propofol-induced sedation because (i) the EEG is consistent during constant low/moderate infusion doses, (ii) propofol causes prominent EEG alpha band activity , and (iii) there is no EEG contamination by surgical artifacts in this post-operative patient.
We started the EEG recording with the 30 mm/s time-scale. We then changed the time-scale to 15 mm/s and started our stepwise protocol with an initial amplitude setting of 1 µV /mm. We changed the amplitude resolution, approximately every two minutes, in a stepwise fashion (µV/mm in eight steps;1→2→3→5→10→25→50→100), while keeping the feed constant at 15 mm/s. After completing the eight steps we changed the feed to 30 mm/s and reversed the amplitude setting from 100 µV/mm to 1 µV/mm in the stepwise fashion described before for another eight steps.
This stepwise protocol is also presented in Fig. 1. We then exported this recording as an EDF file and analyzed the EEG with MATLAB R2017b (The MathWorks, Natick, MA, USA). We based our analyses on the sample rate indicated in the header of the EDF file, which in our case was 178 Hz, because we started with the 30 mm/s feed. For the analyses, we used channel L1, i.e., the signal recorded from position Fp1.
Simulated EEG recording
We further replayed a simulated EEG trace to the SEDLine using an EEG player  to channel L1 (Fp1) and channel L2 (F7). Therefore, we generated a white noise signal using the MATLAB rand function and filtered the signal to the 9–11 Hz range using the MATLAB filtfilt function. We then conducted a stepwise replay of the signal starting with a 30 mm/s feed (i.e., a fs=178 Hz) and a 1 µV/mm amplitude resolution. We increased the amplitude setting every 30 s, changed to a feed of 15 mm/s after 30 s in the 100 µV/mm setting. Then we decreased the amplitude resolution again in the same manner. For the analyses, we used channel L1, i.e., the signal recorded from position Fp1. For the evaluation of clipping, we used the signals from Fp1 and F7 in order to evaluate differences.
Analysis of the EEG stored in the EDF file
In order to examine possible changes in the EEG recording induced by changes in the display settings we calculated the power spectral density (PSD) of the EEG for overlapping EEG segments of 5 s length with a shift of 1 s. Therefore, we used the MATLAB pwelch function with default settings and a NFFT = 512. We also calculated the normalized PSD (nPSD) by dividing the PSD for each 5 s EEG episode by the sum of power in the 0.4 to 30 Hz range. We also used a custom MATLAB function to evaluate the amount of clipping in the recorded EEG for each feed and amplitude display setting. Clipping is a signal distortion that limits measures of the EEG amplitude to a specific maximum value, i.e., it causes horizontal EEG traces. This will interfere with frequency analysis. For our analyses we defined a rather conservative clipping setting by only considering a signal clipped, if it showed the horizontal line for longer than 10 data points. We present the results from the (n)PSD analysis as DSA heatmaps.