-
Notifications
You must be signed in to change notification settings - Fork 2
/
semesterproject.qmd
83 lines (51 loc) · 5.5 KB
/
semesterproject.qmd
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
# 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](grading.md). 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. Thererfore, you will choose an EEG paper (with data) to reproduce yourself.
::: callout-important
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!
:::
What follows is a list of datasets with accompanying papers that I have screened. You can select one of these if you wish, but also feel free to look for your own datasets either on [menar.org](https://nemar.org) or on [openneuro.org](https://openneuro.org). Try to look for datasets below 20Gb of space, with a paper accompanied. Please send a potential candidate to me asap, so I can screen it.
| Subjects | task | analysis | link |
|------------------|------------------|------------------|------------------|
| 20 | grating | decoding | [dataset](https://openneuro.org/datasets/ds004043/versions/1.1.0) |
| 19 | reaching | ERP | [dataset](https://openneuro.org/datasets/ds003846/versions/1.0.1) |
| 24 | pattern symmetry | ERP | [dataset](https://nemar.org/dataexplorer/detail?dataset_id=ds004347) |
| 12 | reward | ERP | [dataset](https://nemar.org/dataexplorer/detail?dataset_id=ds004147) |
| 18 | walking oddall | ERP | [dataset](https://nemar.org/dataexplorer/detail?dataset_id=ds004033) |
| 47 | social task | timefreq | [dataset](https://nemar.org/dataexplorer/detail?dataset_id=ds003702) |
| 17 | game-playing | ERP | [dataset](https://nemar.org/dataexplorer/detail?dataset_id=ds003517) |
## 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! Don't 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](https://mne.tools/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)
::: {.callout-note icon="false"}
## What is the scope of the project report?
The project should not be a dissertation. If you report goes beyond 40 pages (which depending on the amount of plots can easily happen), you should really think whether you need more pages to make your point. A project report with 20 pages or less is completely fine, as long as you discuss what you are seeing / plotting / doing.
:::
::: {.callout-note icon="false"}
## Am I allowed to use other packages
Yes! Everything goes this time around. I do recommend to stick mostly to mne-python though
:::
::: {.callout-note icon="false"}
## Should I use a single jupyternotebook?
No. It is a good idea to use a notebook to display the data / report / results. But one large notebook quickly becomes messy. I'd recommend to encapsulate code in functions and put them in separate files that you can import and use in your notebook.
:::
## 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 31.03.2025