Dear Georgios,

Thank you again for your very helpful answer.
I have succeeded in parcellating my results using the Yeo et al. atlas, with 
the *.dlabel.nii file provided with the parcellated HCP MEG data (107 nodes), 
and with the wb_command -cifti-parcellate line code.

However, I have troubles to find equivalent *.dlabel.nii files for other 
parcellations scheme, like for the example the Desikan-Killiany atlas. Could 
you help me? Also, should I just use the ft_resampledata Fieldtrip function in 
order to go from the 165K or 32K to the 4K dlabel files, or am I missing 



De : Georgios Michalareas <>
Envoyé : lundi 26 mars 2018 19:20
À : Benjamin Chiêm <>;
Objet : Re: [HCP-Users] MEG processing pipelines

Hi Benjamin,

Sorry for late reply but I was out of office.

Regarding single trial source analysis I have put together some code that shows 
you how to project all motor trials for LH into source space in a matrix with 

Nsources * Ntimespoints * Ntrials

beware this matrix for the specific subject is 11GB.

Please find the code attached in file


I have put some comments i hope they help.

Regarding parcellating the data, you have to use the workbench command tool to 
donwload all parcelations from thw 165 K representation to the 4K one used for 
the MEG source level analysis.

I ll have a look and come back to you on this.

Till then I hope the code I am sending you helps



On 3/16/2018 3:38 PM, Benjamin Chiêm wrote:
Dear HCP experts,

In the context of my PhD thesis (using HCP data), I had some questions about 
MEG processing pipelines.
In my research, I somehow need the MEG data during motor task, in the source 
space (i.e. after beamforming), for each trial and each individual separately. 
It seems that the MEG pipelines average data over trials for each individual, 
and that they reduce the temporal resolution of the data when going from 
'tmegpreproc' files to 'srcavglcmv'.

So my two questions are:

- Is it possible, from the 'tmegpreproc' file for each individual, to perform 
sources reconstruction with Linearly Constrained Minimum Variance beamformer 
for each trial independently, and to keep the original temporal resolution 
(about 500Hz)? What are the steps to follow? I tried to play with options like 
« cfg.rawtrial = yes » or « cfg.keeptrials = yes » in ft_timelockanalysis and 
ft_sourceanalysis of the hcp_srcavglcmv_contrasts.m file, but to be honest I am 
not sure of what I am doing...

- After projection on the sources space, how can I parcellate the data, using 
for example the Desikan-Killiany atlas?

I hope I am clear enough in my explanations. Thank you again for the amazing 
work you're doing!


Benjamin Chiêm

HCP-Users mailing list<>





Dr. Georgios Michalareas

Neuroscience Department

Max Planck Institute for Empirical Aesthetics


phone: +49 69 8300479-325


HCP-Users mailing list

Reply via email to