April 5, 2018 at 2:31 am #21701
BRAPH seems like a great toolbox–very straightforward and intuitive, but I have a question about setting up my cohort.
Is it possible to set up an fmri cohort by entering correlation matrices for each subject instead of their ROI timeseries data? And if so, would BRAPH accept Fisher Z-transformed coefficients (some of which could be greater than 1) or would it be best to enter non-transformed r values?
JeffApril 10, 2018 at 10:41 am #21761
Thank you for your interest in Braph. Hopefully, you will find the software useful for your research.
Currently, Braph does not allow for direct input of the connectivity matrices. However, you could modify few lines of the code in order to be able to input the connectivity values instead of time series. We discussed this problem previously on this forum, and I provided a series of steps that accomplish your purpose. You can find the detailed explanation at: http://braph.org/forums/topic/eeg-graph-analysis/ .
With regard to your second question, Braph will accept Fisher Z-transformed coefficients and the fact that some can be greater than 1 will not pose any problem. After you have performed the steps detailed above and inputted your matrices, then you can use the other functionalities of the software without any problem.
I hope that you find this helpful. If you have any other questions, for this topic or anything else, please do not hesitate to contact me and I will do my best to try and help you.
MiteApril 19, 2018 at 3:27 pm #21858
Thank you for your response. Your solution worked great–I’m now able to directly enter correlation matrices.
I do have another question for you though: is it possible to use a density- (or significance-) based threshold but retain the weights of all edges that remain after thresholding?
Thank you again for your help.
JeffMay 2, 2018 at 8:18 am #21930
I am very glad that it worked.
Regarding your question, in its current form, the binarization of the matrices is done by assigning either 1 or 0 to the appropriate edges. For your purposes, I think a slight modification in the code is needed.
I could implement this for you but could you confirm whether this is what you want: You would like to binarize the adjacency matrix at a certain denstiy and then, instead of assigning ones to the edges that remain, you would like to retain their weights. Then you would like to continue your analysis with this weighted matrix.
Could you let me know if this is what you need. Then I could work to implement that and let you know about the result.
MiteMay 2, 2018 at 6:21 pm #21933
Thank you very much for your response. What you described is exactly the option I’m looking for. However, this isn’t something I need right away, as I have continued analyzing my data using weighted undirected matrices. Still, if at some point you can implement that, I am sure it would be very useful.
JeffJune 1, 2018 at 1:22 pm #22124
I implemented this binarization method in Braph in case you might need it in your analysis. In order to do this, the Braph code needs to be changed in the following way:
1) In Graph.m file, you would need to change lines 1606 and 1607 from B = zeros(size(A));
B(A>threshold) = 1;
B = A;
B(A<threshold) = 0;
In this case, the binarization method is modified, so this change will work for all imaging modality options (MRI, fMRI, PET and EEG). You could test this by moving the threshold slider (for example in the fMRI Graph Analysis interface) and verify that it indeed gives the right results.
2) When the binary matrices are plot, the bone colormap of Matlab is used which gives black for zero and white for 1. However, if you would like to change this, you need to modify line 1768 of Graph.m. In order to get equivalence between the weighted and binary matrices, you need to change the colormap from bone to jet. Please note however, that the same weights will be represented in slightly different color tones in both representation because of normalization issues (colors are assigned within a range and the ranges in weighted and binary matrices will not be the same).
I hope that you find this useful in your analysis. If there is anything I can do to help, please let me know.