EEG Semester Project
The goal of this project is to re-analyse and document a real EEG dataset using MNE-Python.
How I will grade the projects can be seen in grading. I value line of thinking and documented motivation much higher than actual results. I want to see that you understand what you are doing.
Which Datasets?
In previous iterations, we took a well understood and standardized EEG dataset, which made the task a bit unrealistic and, frankly, a bit boring. This year, we will try something new. You will choose a paper (with data) to reproduce.
This has the consequence that there will be many (individual) unforseen problems, inconsistencies and problems. This is to be expected and very typical for EEG analyses. See it as part of the challenge!
This is a list of datasets with accompanying papers that I have screened. Please feel free to look for your own datasets either on menar.org or on openneuro.org. Try to look for datasets below 20Gb of space. There is time allocated to discuss the dataset/paper with me.
| Subjects | task | analysis | link |
|---|---|---|---|
| 20 | grating | decoding | dataset |
| 19 | reaching | ERP | dataset |
| 24 | pattern symmetry | ERP | dataset |
| 12 | reward | ERP | dataset |
| 18 | walking oddall | ERP | dataset |
| 47 | (maybe exclude?) social task | timefreq | dataset |
| 40 | music-chords oddball | ERP | dataset |
| 17 | game-playing | ERP | dataset |
What should the project contain?
The idea is to reproduce, but not with a direct reproduction. That is, we will not try to re-do their exact pipeline step-by-step, but rather we will try to obtain their result, with a different pipeline, checking for the robustness of the original analysis pipeline.
A list of the typical steps you will perform: - Preprocessing: Filtering, re-referencening, ICA, Event-handling - (automatic) data cleaning: Time, channel and subjects
Group work
Given the new nature of the datasets and the unforeseeable complications, I propose student groups of 2-3 students.
Where do I get the data?
You can find the link above
Tipps
- Please work modularly! Donβt put all your functions in a single jupyter-notebook. Please use notebooks for your analysis & your functions to a ./src/xyz.py files
- Make use of the BIDS structure and mne-bids-pipeline.
- Thoroughly think about what you are doing and why. And note that down - most people can blindly apply a pipeline, I want to see your reasoning why you included / removed certain steps.
How can I get help / advice?
- Write in the ILIAS Forum. Either others will help, or I myself will answer questions.
- I highly recommend watching the first talk in this session: https://www.crowdcast.io/e/live-meeg-2020/7 by Marijn van Vliet for an fitting introduction of analysing multiple subjects with MNE in a reproducible and documented way. After that, use mne-bids-pipeline as suggested before.
- The MNE-documentation is quite extensive. It is worth looking into. You can also try stackexchange or neurostar for help.
FAQ (will be updated)
Format of report
You can share a git of your project and a report, or the code and the report via Illias. The git can be on github or on the university gitlab. The report can be a jupyter notebook if you prefer to intermingle code and documentation. The report is there to document your decisions and thoughts along the pipeline. I am especially interested why you chose certain steps / parameters / analyses / visualizations. Make grading easy for me and show that you understood what you are doing and what your results are.
You make my life easier if you name the git / zip file with your last name, e.g. Muller_SS2021_EEGSemesterProject.zip, and the folder containing in the zip as well.
Deadline
Please hand in the documents until 31.03.2023