The End Of Knowledge - Vault 3 - GTEC BCI & Neurotechnology Spring School 2024 - Day 1

graph LR classDef gtec fill:#f9d4d4, font-weight:bold, font-size:14px; classDef tools fill:#d4f9d4, font-weight:bold, font-size:14px; classDef demo fill:#d4d4f9, font-weight:bold, font-size:14px; classDef clinical fill:#f9f9d4, font-weight:bold, font-size:14px; classDef research fill:#f9d4f9, font-weight:bold, font-size:14px; A[Martin Walchshofer] --> B[Walkshofer introduces g.tec
biosignal suite 1] A --> C[Suite: auto-updating
research tools 2] C --> D[Tools: simulator, recorder,
MATLAB utilities 3] C --> E[APIs: C#, Python,
MATLAB, licenses 4] A --> F[Demo: setup, acquisition,
processing, analysis 5] F --> G[Cap setup on
colleague Michael 6] F --> H[Simulink model acquires
EEG data 7] H --> I[Filters remove noise,
limit range 8] I --> J[Filtered data shows
blinks, artifacts 9] I --> K[Spectral analysis reveals
alpha peak 10] H --> L[Single channel band
power extraction 11] L --> M[Scope plots extracted
band powers 12] F --> N[Raw data logged
to file 13] F --> O[Streaming layer sends
data out 14] O --> P[Console app receives
EEG stream 15] F --> Q[Offline analysis in
gBS Analyze 16] Q --> R[Bandpass filter cleans
loaded data 17] Q --> S[Trimming, scaling for
EEG visibility 18] Q --> T[Notch filter, bad
channel flagging 19] A --> U[gRecorder: easy clinical
data handling 20] U --> V[Filters clean noisy
gRecorder data 21] V --> W[Blinks, artifacts visible
in gRecorder 22] A --> X[Parallel gHIsys research
data acquisition 23] X --> Y[gHIsys filters, visualizes
both streams 24] Y --> Z[Clean research data
alongside clinical 25] X --> AA[gHIsys streams out
for VR 26] A --> AB[Bluetooth limits g.tec
device streaming 27] A --> AC[Capacitive EEG faces
electrode challenges 28] U --> AD[gRecorder data can
stream out 29] A --> AE[g.tec joins many
research collaborations 30] class A,B,AB,AC,AE gtec; class C,D,E tools; class F,G,H,I,J,K,L,M,N,O,P,Q,R,S,T demo; class U,V,W,AD clinical; class X,Y,Z,AA research;

Resume:

1.-Martin Walkshofer, a software developer at g.tec, introduces the g.tec suite for handling biosignal amplifiers, real-time processing, and offline analysis.

2.-The g.tec suite includes various research tools that can be automatically updated, some requiring licenses and others freely available.

3.-Tools include a functional electrical simulator, recorder for medical use, montage creator, MATLAB tools for data conversion and offline analysis.

4.-APIs are available for C#, .NET, Python, MATLAB and visual stimuli tools. Licenses activate/deactivate access to certain professional features.

5.-Martin will demonstrate setting up an experiment, data acquisition, pre-processing, feature extraction, streaming data between applications, and offline analysis.

6.-Using a gel-based cap with dry pin electrodes on his colleague Michael, Martin starts a new Simulink model in MATLAB.

7.-He adds an amplifier block to acquire data from the connected cap and visualizes the raw EEG signal in a scope.

8.-Notch filters at 50Hz and 60Hz remove power line noise. A bandpass filter from 2-30Hz limits data to the EEG range.

9.-A second scope shows the filtered EEG data. Blinks, clenched teeth, and alpha waves with eyes closed are visible.

10.-Spectral analysis reveals an alpha peak around 10Hz. Setting up data acquisition and basic pre-processing took about 10 minutes.

11.-Martin splits out a single channel and calculates band power features for delta, theta, alpha and beta frequency ranges.

12.-A third scope plots the extracted band powers. Alpha power visibly increases when the subject closes their eyes.

13.-Raw data is logged to a file. The entire setup for acquisition, pre-processing, feature extraction and recording took about 15 minutes.

14.-A Lab Streaming Layer interface is added to stream the raw EEG data out to another application for further processing.

15.-A simple console application is shown receiving the EEG stream to demonstrate separating backend acquisition/processing from frontend GUI.

16.-Next, Martin demonstrates loading the saved EEG file into the gBS Analyze offline processing MATLAB toolbox.

17.-In gBS Analyze, he removes the timestamp channel, designs and applies a 2-30Hz Butterworth bandpass filter to clean the data.

18.-The first 10 seconds containing filter artifacts are cut out. Amplitude is scaled to Â±100 Î¼V for good EEG visibility.

19.-A 50Hz notch filter further cleans noise on one channel. Signal quality metrics flag a couple potentially bad channels.

20.-In a clinical setting, the compiled gRecorder application allows easy data visualization, filtering and recording without any programming.

21.-Connecting to the amplifier in gRecorder, raw unfiltered EEG is very noisy. Applying a 50Hz notch and 2-30Hz bandpass cleans it.

22.-Eye blinks, muscle artifacts, and alpha waves are visible in the filtered clinical-grade EEG. Data can then be saved to disk.

23.-While recording in gRecorder, a parallel research experiment is set up in gHIsys using its remote data acquisition block.

24.-The gHIsys model applies filters and visualizes both the raw gRecorder data and additionally filtered research-grade EEG.

25.-Eye blinks, EMG, and alpha are visible in both streams, with cleaner research-grade data. This parallel setup took just minutes.

26.-gHIsys models cannot directly export to embedded systems, but can stream out data to combine with e.g. VR in other applications.

27.-Up to 2 g.tec devices may stream on one PC before Bluetooth limits are hit. Multiple PCs can merge more streams.

28.-Capacitive EEG still faces challenges with electrode size, interference, and shielding compared to gel-based electrodes in close skin contact.

29.-While no open SDK exists for custom hardware in gRecorder, its data can stream out to other applications via API.

30.-g.tec actively participates in around 15 international research projects at any given time with about 100 total partners.

Knowledge Vault built byDavid Vivancos 2024