Knowledge Vault 3/90 - G.TEC BCI & Neurotechnology Spring School 2024 - Day 9
Off-line analysis of ERPs with g.Bsanalyze
Rupert Ortner, g.tec medical engineering GmbH (AT)
<Resume Image >

Concept Graph & Resume using Claude 3 Opus | Chat GPT4 | Llama 3:

graph LR classDef main fill:#f9d4d4, font-weight:bold, font-size:14px; classDef analyze fill:#d4f9d4, font-weight:bold, font-size:14px; classDef test fill:#d4d4f9, font-weight:bold, font-size:14px; classDef mindbeagle fill:#f9f9d4, font-weight:bold, font-size:14px; classDef process fill:#f9d4f9, font-weight:bold, font-size:14px; Main[Rupert Ortner] --> Analyze[Offline ERP analysis
with GBS Analyze. 1] Main --> Test[Simple demo using
test data. 2] Test --> Data[8 EEG channels,
10 trials. 3] Test --> Visualize[Load, zoom, scroll,
visualize attributes. 4] Test --> Average[Baseline correction, separate
class averages, stats. 5] Average --> Plot[Plot target/non-target ERPs,
highlight differences. 6] Main --> MindBeagle[Analysis of MindBeagle
device data. 7] MindBeagle --> Import[Import unfiltered data
into GBS Analyze. 8] MindBeagle --> Filter[Bandpass 0.5-30 Hz,
notch 50/60 Hz. 9] MindBeagle --> Adjust[Adjust scaling, exclude
channels, load geometry. 10] MindBeagle --> Trigger[Cut EEG into
trials using markers. 11] Trigger --> Define[Define target/non-target triggers,
baseline, post-stimulus. 12] MindBeagle --> Artifact[Average reveals artifacts
needing correction. 13] Artifact --> Detrend[Apply detrend to
each trial. 14] Artifact --> Reject[Manual artifact rejection,
exclude bad trials. 15] MindBeagle --> Clean[Final averages cleaner
after processing. 16] Main --> Options[Change scales, select
channels/trials, hardware triggers. 17] Main --> KeySteps[Filtering, triggering, artifacts,
averaging are key. 18] Main --> Learn[Test data for
learning software. 19] Main --> RealData[Real data needs
careful preprocessing. 19] Main --> GUI[GBS Analyze: interactive
GUI for EEG/ERP. 20] class Main main; class Analyze,Average,Plot analyze; class Test,Data,Visualize test; class MindBeagle,Import,Filter,Adjust,Trigger,Define,Artifact,Detrend,Reject,Clean mindbeagle; class Options,KeySteps,Learn,RealData,GUI process;


1.- Rupert demonstrates how to do offline analysis of event-related potentials (ERPs) using GBS Analyze MATLAB-based software.

2.- Rupert starts with a simple demo using test data included with GBS Analyze to demonstrate averaging and statistical analysis of ERPs.

3.- The test data has 8 EEG channels, 10 trials, and is already triggered, containing target and non-target trials from a visual paradigm.

4.- Rupert shows how to load the data, zoom in/out, scroll through trials, and visualize trial and channel attributes.

5.- In the Analyze > Average menu, options are available for baseline correction, computing separate averages for different trial classes, and statistical analysis.

6.- Rupert demonstrates plotting the average target and non-target ERPs for each channel, with significant differences between classes highlighted.

7.- Next, Rupert shows analysis of data from their new MindBeagle device, importing the data into GBS Analyze.

8.- MindBeagle data is stored unfiltered to preserve information. Rupert shows how to bandpass filter the EEG channels from 0.5-30 Hz.

9.- A notch filter at 50 Hz (Europe) or 60 Hz (US) is used to remove powerline interference. Channels are excluded as needed.

10.- Scaling is adjusted to visualize the EEG signals. A geometry file specifying 3D positions of electrodes can be loaded for topographic plotting.

11.- Triggering is performed to cut the continuous EEG into segments (trials) based on stimulus events, using markers (software triggers).

12.- Target and non-target triggers are defined, specifying a 100 ms pre-stimulus baseline and 700 ms post-stimulus, resulting in 270 trials.

13.- Averaging is performed again, revealing some artifacts in certain channels/trials that require correction.

14.- One quick option to handle artifacts is applying a detrend to each trial to remove drift.

15.- Manual artifact rejection is also shown, scrolling through trials and marking artifactual ones, which are then excluded from averaging.

16.- The final averages look cleaner after detrending and artifact rejection.

17.- Analysis options like changing scales, selecting subsets of channels/trials, and using hardware-based triggers are mentioned.

18.- Filtering, triggering, artifact handling and averaging are key steps in ERP analysis.

19.- Test data allows learning the software, while real data requires careful inspection and preprocessing.

20.- GBS Analyze provides an interactive GUI to flexibly load, preprocess, analyze and visualize EEG/ERP data.

Knowledge Vault built byDavid Vivancos 2024