Thanks for explaining, and thanks for updating example 10. When I follow those 
commands to get an average vertex area, unfortunately, the palm command 
ultimately fails with the error:

Error using palm_miscread (line 157)

Not enough input arguments.

Error in palm_takeargs (line 1509)

            plm.srfarea{s} = palm_miscread({s});

Error in palm_core (line 33)

[opts,plm] = palm_takeargs(varargin{:});

Error in palm (line 81)


Any thought why this may be? Should I attach the file to verify that it is the 
correct output for this?

It runs smoothly if I just use one input, such as from the HCP data. Is that the 
file you would suggest using (for the left hemisphere)? I'd prefer to use a 
local index of vertex area if I can as that seems more appropriate. Do you 
think that is important?

As for the previous errors, I had left out the -ise flag, so I'm not sure what 
the program was putting out (I guess nothing).

Thank you again for all your help,
From: <> on behalf of 
Anderson M. Winkler <>
Sent: Friday, September 16, 2016 11:33:51 AM
To: Michael F.W. Dreyfuss
Subject: Re: [HCP-Users] Palm with TFCE on Cifti data

Hi Michael,

This isn't correct I'm afraid. Even if PALM run to completion, with these 
inputs it is shuffling the timeseries data within-subject, but because of 
serial correlations, this violates the most important assumption of permutation 
tests: the one of exchangeability.

The analysis needs to use with -i input file(s) that include all subjects. The 
reason why PALM allows multiple inputs is that one can test multiple different 
imaging modalities, such as cortical area and BOLD FMRI, and/or different 
variants of the same modality, such as different ICA maps, or different 
diffusion tensor metrics, or different morphological metrics, or different FMRI 
tasks. The -i isn't meant to be used one per subject.

The -s should be one per input, and should be the group mesh to which all 
subjects were resampled, as opposed to one per subject.

All the best,


PS: I the last email I accidentally clicked "reply to" instead of "reply to 
all", only to more than an hour later realise it hadn't been sent to the HCP 
list. Ideally we should keep conversations as this in the list as it can be 
helpful to others in the future. Hope you won't mind that I'm cc'ing the list 

On 16 September 2016 at 13:07, Michael F.W. Dreyfuss 
<<>> wrote:

Thank you very much.

Sorry,  I don't think I was clear about what I was suggesting. I was proposing 
using the midthickness surface file generated for each participant rather than 
an average in separate -s arguments, so something like:

palm -i -i -i {...} -s -s -s {...}

such that each subject's input file is matched by their own corrresponding 
midthickness surface file. I tried this and palm was able to run. Would you 
recommend creating a group average or using the average from the HCP data over 
this approach? If you recommend a local group average, what would be the 
wb_command to do that on the midthickness data?

Sorry, I wasn't sure what you were referring to about compiling and skipping 
the conversion. What is that in reference to?

Thank you again,


<<>> on behalf of 
Anderson M. Winkler <<>>
Sent: Friday, September 16, 2016 4:19:06 AM
To: Michael F.W. Dreyfuss

Subject: Re: [HCP-Users] Palm with TFCE on Cifti data

Hi Michael,

The "-s" option can take 1 or 2 arguments. The first is mandatory and it is a 
surface file (.surf.gii for instance). If only this argument is given, PALM 
will use this file and do two things with it: it will use it for the adjacency 
information and it will also use this file to calculate the area-per-vertex, 
which is necessary to compute the cluster extent statistic, and also to 
calculate the area of the support region of TFCE. This means that if only one 
argument is given, in order to have a reasonable approximation, this file 
should look like a brain (e.g., midthickness or white, but not sphere).

If a second argument is given, this second argument should be a scalar file 
with the area per vertex for the same vertices of the surface file. Ideally 
this file should be an average of the areas of the same subjects in your 
sample, but if in your lab you are having difficulties in calculating it, use 
the midthickness_va.shape.gii that Tim suggested. It is based on a different 
set subjects, but should already be better than using the area of the surface, 
even if that looks like a brain already. If the area file is supplied (i.e., 
the 2nd argument then), then the surface file (1st argument) can have any 
shape, even a sphere then.

The option -s should be used for each -i. If you enter left and right 
separately as inputs, then there should be two -s options, one for left and 
another for right, in the same order.

About compiling: please give it a try for your platform -- maybe you can skip 
that step with the conversion then.

Hope this helps.

All the best,


On 15 September 2016 at 21:57, Michael F.W. Dreyfuss 
<<>> wrote:

Ah ok, so it sounds like the best and most data-specific thing to do would be 
to use each subjects midthickness file as separate inputs to -s. I'm referring 
to the file that is produced from the preprocessing stream in 

Is that correct?

Thanks again,


From: Timothy Coalson <<>>
Sent: Thursday, September 15, 2016 4:46:12 PM

To: Michael F.W. Dreyfuss
Cc: Anderson M. Winkler;<>; Dierker, 
Subject: Re: [HCP-Users] Palm with TFCE on Cifti data

TFCE doesn't use distances along the surface, so using spherical surfaces and 
the average vertex area files should give the correct answer (it is best to 
provide average vertex area files when using any group data).  I don't know if 
PALM does anything that uses surface distances (such as smoothing), but for 
operations that do (like wb_command -cifti-smoothing), spherical surfaces are 
not recommended, instead use either individual surfaces (matching the data the 
subject is for), or a group average midthickness surface and average vertex 

As in my previous email (though it was a bit hidden), the group average 
midthickness surfaces should be on connectomedb - open dataset on the HCP 900 
(soon to be more) dataset, then scroll down to "group average".  The PALM 
example talking about TFCE is very new, and needs some revision to be correct.


On Thu, Sep 15, 2016 at 3:21 PM, Michael F.W. Dreyfuss 
<<>> wrote:

OK thank you for your answer and for clarifying. I am still a bit confused 
about how to proceed with what to use for the surface file for the -s option 
with palm for TFCE on the .gii data.

I was wondering because on example 10 for palm it says "Various common meshes 
at different resolutions used in the HCP pipelines are available 
 Use preferably the midthickness." Where "here" is a link to :<>

So I was hoping that the 32k midthickness files would be usable fore the -s 
option of palm there for processing surface data, but I guess not? Are there 
files recommended for this?

Thank you,


From: Timothy Coalson <<>>
Sent: Thursday, September 15, 2016 4:05:19 PM
To: Michael F.W. Dreyfuss
Cc: Anderson M. Winkler;<>; Dierker, 

Subject: Re: [HCP-Users] Palm with TFCE on Cifti data

That .shape.gii file is not a surface file, you need .surf.gii files to use 
that command.  The standard_mesh_atlases folder does not contain any anatomical 
surfaces (it only has spheres).

See wb_command -gifti-help for an explanation of the types of gifti file:<>

The "midthickness_va" .shape.gii files in the resample_fsaverage folder are the 
vertex areas of the S900 group of subjects, averaged across subjects, so that 
documentation needs to be changed (the group average midthickness *surfaces* 
are not here, try connectomedb, "open dataset" on the main HCP project, then 
scroll down to "group average data").  This folder wasn't intended for the 
purposes of processing data, it exists for resampling between atlases.

Side note, I believe octave can handle the default base64_gzip gifti encoding 
if you compile the zip implementation that comes with the gifti toolbox (via 
"make PLATFORM=octave" in @gifti/private).


On Thu, Sep 15, 2016 at 11:25 AM, Michael F.W. Dreyfuss 
<<>> wrote:

Thank you very much! This is very helpful.

Two small problems remain before running palm on surface data:

1) I cannot download the left and right 32k midthickness meshes from the site 
linked to in the examples page:


2) I get errors when I try to get average areas for the -s option. When I use 
-surface-vertex-areas on an individual subjects for example I get the following:


wb_command -surface-vertex-areas data_L.func.gii


ERROR: Number of data arrays MUST be two in a SurfaceFile.

I am again not sure what this error means and I have not found any information 
about it or how to resolve it. All the other steps are super well explained and 
demonstrated, but then the three steps using wb_command on how to find average 
areas of gii files are quickly passed over. Is there any explanation on how 
that might work?

Thank you!


<<>> on behalf of 
Anderson M. Winkler <<>>
Sent: Thursday, September 15, 2016 5:39:12 AM
To: Michael F.W. Dreyfuss
Cc: Dierker, Donna;<>

Subject: Re: [HCP-Users] Palm with TFCE on Cifti data

Hi Michael,

Please see an example in this page:<>

The Example 9 is for simple statistics on brainordinates (no spatial 
statistics) and Example 10 is for TFCE and cluster-level inference.

Thanks to Tim Coalson for the help with the wb_command options.

Hope this helps!

All the best,


On 15 September 2016 at 07:21, Anderson M. Winkler 
<<>> wrote:
Hi Michael,

TFCE needs to be run separately for the surface-based and for the volume-based 
parts of the CIFTI file. This means that the input file has to be separated 
using wb_command, then PALM be executed, then the results need to be corrected 
via Bonferroni or Šidák. I will add hopefully today still a section in the 
documentation with the steps and update here.

Also, although it's unrelated to this error, Donna is right: the extension for 
the -t option wasn't supplied correctly.

All the best,


On 15 September 2016 at 06:22, Michael F.W. Dreyfuss 
<<>> wrote:

Thank you,

The $ContrastName is the argument of the -o option, so it is just the name 
designated for the output. There are no csv files unless you mean the .mat and 
.con files which are the arguments of -d and -t respectively for the design 
files and contrast files. Since it is a group average, it actually runs without 
the contrast file as well.

Again, this whole command runs fine and gives reasonable output with the -fdr 
flag instead of -T, but I would like to be able to run TFCE here. Has the -T 
option been used with palm on cifti data successfully? Maybe it is not meant is 
only for use on volumetric data?

Thanks a lot,


From: Dierker, Donna <<>>
Sent: Wednesday, September 14, 2016 11:09:04 PM
To: Michael F.W. Dreyfuss
Subject: Re: [HCP-Users] Palm with TFCE on Cifti data

Michael, if $ContrastName is a csv file, make sure its extension is csv.  I’ve 
gotten errors like that before when I named it .con.

> On Sep 14, 2016, at 11:29 AM, Michael F.W. Dreyfuss 
> <<>> wrote:


> Hello, I am trying to run palm on cifti data with TFCE to find mean 
> activation across a group. When I use the -T flag, the program crashed with 
> the error:


> Command:

> palm -i ${ContrastName}.dtseries.nii  -transposedata  -d 
> $LevelThreeFEATDir/design.mat  -o $ContrastName -t 
> /tmp/CtxFGNG/CtxFGNG_Level3_Analysis/desion -T -ise


> Error:

> 0.01% [Design 1/1, Contrast 1/1, Shuffling 1/10000, Modality 1/1]

> Saving file: FoodNogo_dat_tstat

> Index exceeds matrix dimensions.


> Error in palm_tfce (line 77)

>     elseif plm.Yisvtx(y) || plm.Yisfac(y),


> Error in palm_core (line 1680)

>                                 Gtfce{y}{m}{c} = 
> tfcefunc(G{y}{m}{c},y,opts,plm);


> Error in palm (line 81)

> palm_core(varargin{:});



> Alternatively, if I use the same command, but replace -T with -fdr, the 
> program runs smoothly, so I do not think the problem is with my design files 
> or data unless they need to be generated in a special way for TFCE using 
> palm. Are there any recommendation or previous examples of how I can use TFCE 
> on cifti data using palm?


> Any help on this is very appreciated.


> Thank you,

> Michael

> _______________________________________________

> HCP-Users mailing list




The materials in this message are private and may contain Protected Healthcare 
Information or other information of a sensitive nature. If you are not the 
intended recipient, be advised that any unauthorized use, disclosure, copying 
or the taking of any action in reliance on the contents of this information is 
strictly prohibited. If you have received this email in error, please 
immediately notify the sender via telephone or return mail.

HCP-Users mailing list<><>

HCP-Users mailing list<><>

HCP-Users mailing list

Reply via email to