Integrate the code from Danielle into the experiment to be able to use first participant

[yacineMahdid]

This is important since we want to be able to validate if the dpli index make sense and we have recovery information about the participant WSAS02 which doesn't have a similar headset than all the other participants.

First task to do:

• [x] Generate a csv file which has the form egi_channels, bp_channels right now we have a third column that has the order since the second column is not ordered properly. (Can be Python with pandas) : the file however is currently useless since the mapping is not consistent between our recording channel location and what is on the csv file, Update 2: After some back and forth Danielle provided me with a file that I now follow for the mapping
• [x] Make a module in MATLAB that will dynamically calculate the euclidean distance between two electrode from two different headsets and select the best label for this channel. This will need to use Danielle Nadin code : Only thing we need to change if we want to is the mapping file bp_to_egi_mapping.csv

Will need to run four analysis

• [x] similarity matrix (first way)
• [x] dpli-dri attempt 2
• [x] similarity matrix (second way)
• [x] dpli-dri attempt 3

These should be documented in the corresponding wikipage

• [x] Restructure the code in the repository to be better documented.
• [x] Document an updated version of the analysis from dpli matrices to similarity matrices. In the documentation explain what the main problems were and what I did to fix them.
• [x] Write a final report on attempt 1,2,3 with all the participants.
• [x] Close this milestone and celebrate :1st_place_medal:

[yacineMahdid]

I've documented the mapping over here in a wiki page

[yacineMahdid]

We are having some issue with the converter as the version I have in NeuroAlgo doesn't allow for channel with name like Fsomething. I will make a conversion module to take each channels that doesn't start with E and map it to its equivalent.

[yacineMahdid]

One of the similarity matrices, done with the first way of doing it

[yacineMahdid]

There is clearly something wrong since we get a a matrix with higher contrast for the alpha vs recovery in the previous picture. Should take a look at the raw matrices.

[yacineMahdid]

Given this dPLI matrices what we have above doesn't make any sense.

[yacineMahdid]

The problem seems to be caused by a mismatch in the regions localization. We see the exact same pattern, but shifted in one case and not in another.

[yacineMahdid]

When using the min-max normalization we do not get the same shift

[yacineMahdid]

The similarity matrix looks like this now

[yacineMahdid]

However we still have larger difference in the baseline vs recovery comparisons

[yacineMahdid]

To validate that the number of channels are making our analysis less robust I'll run the analysis again but with only the left brain for all other participant than WSAS02

[yacineMahdid]

The main issue was in the filter_matrix function which was not filtering the matrices properly. I also added normalization as Dr. Blain-Moraes suggested, which helped :+1:

Here is the results!

[yacineMahdid]

I will move away from Github Wiki and instead directly use markdown files inside the github repository. This will be easier to structure and will allow for the documentation to match up with the codebase current state.