Welcome to nirs-toolbox Discussions!

[huppertt]

👋 Welcome!

We’re using Discussions as a place to connect with other members of our community. We hope that you:

• Ask questions you’re wondering about.
• Share ideas.
• Engage with other community members.
• Welcome others and are open-minded. Remember that this is a community we
  build together 💪.

To get started, comment below with an introduction of yourself and tell us about what you do with this community.

[LeiaHealy]

Hi Dr. Huppert!

I just have a quick question regarding the differences between the probe 1020 and probe class. When performing image reconstruction, everything works fine until I run the ImageReconMFX module in which the below error message pops up. I think that the issue may be that my SubjStats contains a probe of the 1020 class, so its probe.link variable is different from the standard probe.link variable. Is there anyway to convert from one to the other? Or am I doing something wrong in the subject analysis portion to have my probe.link end up being the inherited probe1020 class rather then just a probe?

################
Error Message:

Error using tabular/ismember (line 37)
A and B must contain the same variables.

Error in nirs.modules.ImageReconMFX/runThis (line 114)
[ia,ib]=ismember(thisprobe.link,S(idx).probe.link,'rows'); <----- comparing probe1020.link to probe.link

Error in nirs.modules.AbstractModule/run (line 77)
out = obj.runThis( inputs );
################

Best,
Liam Healy

[xuetongzhai]

Hi Liam. I am a postdoc working in Dr. Huppert's lab. Can you send me your data and code to my email [email protected]. Then I can take a look and let you know.

-Xuetong

[huppertt]

My best guess is that the stimulus you are creating is a struct and not the actual StimulusEvents class.

> raw=nirs.testing.simARNoise; > stim=nirs.design.StimulusEvents; % create empty stim class. <<<< I THINK YOU ARE MISSING THIS LINE > stim.onset=[30:30:300]; % events at 30,60,90…300s > stim.dur=10*ones(1,length(stim.onset)); % all events are 10s duration > stim.amp=ones(1,length(stim.onset)); > raw.stimulus('MyTaskName')=stim; % assign to the dictionary as “MyTaskName” as the condition > raw.draw; % this should now show the task marks > stim.draw; % this should draw the event marks

My postdoc Hendrik (cc’d) can help if this doesn’t solve the issue for you.

…

-Ted Theodore Huppert, PhD University of Pittsburgh Department of Electrical and Computer Engineering Email: ***@***.******@***.***> Phone: 1-412-383-5167 Website: www.huppertlab.net From: FatimaSiba ***@***.***> Reply-To: huppertt/nirs-toolbox ***@***.***> Date: Tuesday, August 24, 2021 at 11:44 AM To: huppertt/nirs-toolbox ***@***.***> Cc: huppertt ***@***.***>, Author ***@***.***> Subject: Re: [huppertt/nirs-toolbox] Welcome to nirs-toolbox Discussions! (#2) Dear Dr. Huppertt and lab team members I'm trying to process some practice for my experiment. I'm still in the beginning of the process. I'm editing the stimulus by creating my own dictionary class variable and assigning it to the raw data. When I visualize the data the stimulus looks correct. However when I try running the GLM model, I'm running into an error: Warning: No stim conditions left. Did you provide the correct stim names? Error using fields Invalid input argument of type 'double'. Input must be a structure or a Java or COM object. Error in nirs.modules.GLM/runThis (line 154) if(~ismember('regressor_no_interest',fields(st))) Error in nirs.modules.AbstractModule/run (line 77) out = obj.runThis( inputs ); I could see in the GLM function the st variable is empty and that's why the error is occurring. However, I'm not sure why that is. The stimulus names are correct in the raw.stimulus.key as well as raw.stimulus.values.names. I appreciate your help Thanks Fatima Sibaii — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<#2 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/ALNU3OXOXALSXCXH5A4WRODT6PZCXANCNFSM4VFTEJPQ>. Triage notifications on the go with GitHub Mobile for iOS<https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675> or Android<https://play.google.com/store/apps/details?id=com.github.android&utm_campaign=notification-email>.

[FatimaSiba]

Hey Dr. Huppertt
Thank you, we figured out the issue and it is now resolved.
Fatima Sibaii

[bbreeze25]

dear all, what version of Matlab would be the most stable to work with analyzIR? looks like I am getting errors that are Matlab version specific

[FatimaSiba]

Hello

I have found a post from 2018 that discussed getting the residuals after running the GLM using advanded.nirs.modules.GLMResiduals.
Here is the link:
https://huppertlab.net/discussion/forum/topic/can-residuals-be-extracted-after-glm/
I have tried using this function, however I'm getting this error:
Unable to resolve the name nirs.modules.GLMResiduals.
I'm not sure what that error means.
Also, if this were to work, are the residuals from the GLM considered "filtered time series"?

Thanks
Fatima

[huppertt]

I think the code was moved to advanced.nirs.modules.GLMresiduals. The advanced namespace contains a bunch of stuff that is either work-in-progress or was part of a paper comparing methods and determined to be not the best method (e.g. some of the variations on how to use short-separation channels are in there after we found putting them in the GLM was the best approach). The time-course data that will come back after this module will have the average response removed but hasn’t been really filtered beyond that unless you additionally have correction regressors in the GLM (e.g. short-separations or DCT/polynomial detrending terms). E.g. residual = Y – X*beta, so if X has “filtering” regressors, these will also be projected out of the residual. Beta is estimated by the same AR-IRLS GLM model, but it goes back to the “raw” Y to compute the residuals once it finds beta Theodore Huppert, PhD Associate Professor University of Pittsburgh Department of Electrical and Computer Engineering Email: ***@***.******@***.***> Phone: 1-412-383-5167 Website: www.huppertlab.net<http://www.huppertlab.net> Schedule a meeting with me at https://calendly.com/huppert1 From: Fatima Sibaii ***@***.***> Date: Tuesday, February 22, 2022 at 12:18 AM To: huppertt/nirs-toolbox ***@***.***> Cc: huppertt ***@***.***>, Author ***@***.***> Subject: Re: [huppertt/nirs-toolbox] Welcome to nirs-toolbox Discussions! (Discussion #2) Hello I have found a post from 2018 that discussed getting the residuals after running the GLM using advanded.nirs.modules.GLMResiduals. Here is the link: https://huppertlab.net/discussion/forum/topic/can-residuals-be-extracted-after-glm/ I have tried using this function, however I'm getting this error: Unable to resolve the name nirs.modules.GLMResiduals. I'm not sure what that error means. Also, if this were to work, are the residuals from the GLM considered "filtered time series"? Thanks Fatima — Reply to this email directly, view it on GitHub<#2 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/ALNU3OWFUOIHOLBLZNWJBITU4MMB5ANCNFSM4VFTEJPQ>. Triage notifications on the go with GitHub Mobile for iOS<https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675> or Android<https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub>. You are receiving this because you authored the thread.Message ID: ***@***.***>

[FatimaSiba]

Dear Dr. Huppert

Thank you so much for your prompt reply! I changed the handle in MATLAB to match the new location of the GLMresiduals function, and it is now running. However, I'm running into an error at this line of code: data(i).data = data(i).data - X * reshape(S(i).beta,[nchan ncond])';
I believe it's because the size of X depends on my conditions so it is of size: datapointsconditions, while beta is reshaped to (numbers of channels(betalength/number of channels)). I have 4 conditions (so X is timepointsx4), but beta has a length of 192, with 64 channels so it is reshaped to 3x64. Therefor, matrix multiplication is not working.
Also, the X created is all 0s, no values.
I appreciate any tips on how to solve this error!

Thank you
Fatima

[DemianVera]

Hi, Dr. Huppert. I was watching this webinar and trying to reproduce the ROC curve analysis. I couldn't find the exact code you used in that webinar, so I managed to do it myself in Matlab (2021a), following the video and using the same data files. For any reason, when I compute the ROC curves, two problems appear: the first one is that there seems to be an error in ChannelStatsRoc.m, in Line 153; I suppose that where it says if(iscellstr(types(i))), it should say if(iscellstr(types(j))). Doing this replacement, it works (it doesn't complain, at least :-/ ).
The second problem I find is that all ROC curves (for any pipeline, and for hbo, hbr and joint) are almost straight lines with an area under the curve of ~0.5.
I'm using a recent version of nirs toolbox (I downloaded it last week from github).

Thanks for your time and help, and sorry for bothering you with this.
I really appreciate all you're doing for the fNIRS community. It's a great, great job.

Best regards from Argentina,

Demián Vera.

[xuetongzhai]

Hi Demián,

Let me look at it and try to get back to you by this weekend.

[DemianVera]

Hi! Sorry for bothering you again; As I see no answer, I think there was no problem with the code and maybe I was wrong. Or maybe you had no time to check it out. Thanks for your time, anyway!

[nicdc]

Just to add, I've had the same issue using this script. It was working fine when I tested it previously (soon after Prof. Huppert gave the webinar at NIRx) but some subsequent change seems to have broken this script.

[nicdc]

Looked into this a bit more, I have a version of the toolbox from 17 months ago, when my TortoiseHg stopped updating the local repo due to some authentication issue. Reruning the same code that was shared in the demo on the same data with two versions of the local repo gives me two very different results.
simulations_17monthsago.pdf
simulations_today.pdf

[DemianVera]

Thanks! Yes, they are very different. Something happened and no one noticed...

[XueruiPeng]

Dear Dr. Huppert,

Thanks a lot for your wonderful and useful toolbox.
I have some questions about parametric modulation analysis.

When I use nirs.design.split_parametric to manipulate my dataset, I have nan values in cond: amp.
Specifically, here is my code to do so:

 stimTable = nirs.createStimulusTable(myRawdata);
 %% then I put amp value of my cond1 into the stimTable with some code %%
 job = nirs.modules.ChangeStimulusInfo();
 job.ChangeTable = stimTable;
 rawChanged = job.run(rawChanged);
rawChanged = nirs.design.split_parametric(rawChanged,'con1*(1+Amp)',true);

After running the code, I got a new stim named con1:amp, with the amp values being all nan. And when stats analysis is performed, this condition is automatically removed. Did I use the function incorrectly?

Then to try to solve this, I manually added the parametric modulation term into my data set. So I created a new cond named ‘con1:amp’ with the same onset and duration with con1 and mean-centered amp value. Then perform the stats. Is it correct to do so?

Furthermore, I have two environment-related values in each trial (the two values are highly correlated, r = 0.73). So when I want to see the effect of one of them, I need to include another one in the model to regress its effect out. Can I add the second parametric modulation as I did for the first one? Then I have con1 (with amp all equal to 1), con1:amp1, con1:amp2. Should I have other steps to deal with their high correlation (such as orthogonalizing them beforehand)? Or, from my understanding, the regression model can help me deal with this kind of colinearity (I use the AR_IRLS model). Does the order of the two parametric modulated terms matter?

I appreciate any answers, tips, and suggestions about that.
Xuerui

[armandosajunior]

Hi guys,

Sorry if this is a silly question. I love the nirs toolbox, I'd love to continue using it, but I don't know how to remove bad channels in the toolbox - which is kind crucial to me. I usually do some preprocessing, glm and then linear mixed model. But if I try to remove bad channels, I usually get some errors, that I think is related to an uneven number of channels (e.g. Unable to perform assignment because the indices on the left side are not compatible with the size
of the right side.

Error in nirs.modules.MixedEffects/runThis (line 225)
X(idx, :) = X;

Error in nirs.modules.AbstractModule/run (line 73)
out = obj.runThis( inputs );)".

Can anybody help me?

[DemianVera]

Hi everyone!
I use nirs toolbox to process my fNIRS data. Now I'm trying to open some data (BIDS) that contains data in snirf format. So I must use the nirs.io.loadSNIRF function; however, a function called snirf2data throws errors when trying to read the field SourcePos (line 60 of the present function). I think this happens because my data has the fields sourcePos3D and detectorPos3D. As I can see, this function takes that into account, but only partially.
Can you help me, please? Is there any other way to load snirf data?

Thanks in advance!

[AlinaSchulte]

Hi!

To identify channels with bad data quality, I previously used the scalp coupling index. In addition, I found the structure noise index (nirs.math.structnoiseindex) that is used within the QA report to identify bad channels. It seems to detect channels with large amplitudes (median deviations) that remain after removing autocorrelations. I'd like to understand a bit better what it is doing exactly. Has anyone else used it and could elaborate on its functioning?

Best regards,
Alina

[AlinaSchulte]

Hi,

I am wondering whether it is okay to use a band-pass filter before applying the AR-IRLS algorithm to my data. Within the code I see that low-pass filtering is not recommended (see screenshot below) and can lead to unstable AR coefficients. I could not figure out what that means for the data and the glm results, but trying my filter on a simulated dataset does not seem to cause visible problems and helps finding the simulated response.

Looking into the literature, I see that in some publications filtering is used with the AR-IRLS glm model estimation. That's also the case for the paper by Santosa et al., 2020. Here however, the reversed problem is mentioned (high-pass filtering being problematic).

Can anyone help with this?

Best,
Alina

[AMGoodwill]

Hi, I am wondering if the connectivity functions in nirs toolbox could be used for task-based as well to examine dynamic networks?
Thank you!
Alicia

[Raya2023]

Hi,

Thank you for sharing your toolbox! @huppertt @xuetongzhai @hendriksantosa
How can we get support for our inquiries or the issues occurs when using NIRS toolbox?
I'm facing a problem when loading my .nirs files, some of them are loaded with missing stimulus table or missing column, although I confirmed that their original csv files (acquired by Hitachi) include the mark values. And some other files are loaded fine (converted using the same script and loaded using the same function as the files with missing stimulus). What would be the solution to get my stimuli values right?
BTW, when I tried loadHitachi function for directly loading Hitachi csv files, MATLAB hangs.

Thank you in advance.

[shatton412]

Hello,
I had a concept question to check my knowledge before I start my analysis. I am interested in doing multimodal FNIRS and EEG data fusion. I know you can convolve the eeg signal with the HRF from the fnirs using eeg.modules.HRFConvolve. However, my data set is a resting data set. Most of the papers I read make it seem that most HRF's can only be extracted from task-based experiments. Am I able to use eeg.modules.HRFConvolve on my resting data? Or is this only feasible on task-related paradigms?
Thank you in advance!

[pimbrouwer]

Hello Dr. Huppert!

Thank you for the NIRS toolbox. I wanted to explore the possibilities and was trying to run the fnirs_example_shortsep.m demo. The dataset which is used in this demo can't be found at huppertlab.net anymore. Could you share the NIRx_data_SS.zip with me?

Thank you!