White matter specific HRF for fMRI analysis?

Hi experts,

I want to examine whether white matter changes during task-fMRI. I used AFNI to analyze the data with the traditional procedure as gray matter. However, the white matter does not show obvious activation. I think this may be due to the hemodynamic response function (HRF). So is there any methods to choose the HRF in AFNI? Or is there any other tools which provides white matter specific HRF for task-fMRI analysis?



Hi @Roger1219

If you also have resting state data, you can use the rsHRF toolbox (GitHub - compneuro-da/rsHRF: rsHRF: A Toolbox for Resting State HRF Deconvolution and Connectivity Analysis (MATLAB)) to derive HRFs in a given part of the brain, as in Anomalous and heterogeneous characteristics of the BOLD hemodynamic response function in white matter | Cerebral Cortex Communications | Oxford Academic. Of note, the HRF is not consistent throughout white matter, which complicates things.

Not sure how to use these outputs to AFNI.


1 Like

Hi, Roger:

I think this paper outlines and discusses methodology for exactly what you would want—estimating a hemodynamic response function (HRF) at each voxel, without assuming a constant/canonical shape (and including regularization):

Both AFNI’s standard TENT functions and the proposed “smooth HRF” functionality provide a lot of benefits over canonical HRFs—I was actually astounded at the variability of HRF across just gray matter. If you want to investigate BOLD response in white matter, well, that is even another huge degree of variability. The “smooth HRF”'s main advantage over TENTs is the regularization aspect, which essentially can be viewed as noise reduction.

The codes related to that paper are available in GitHub, from its page in the AFNI Codex:

Specifically, here are some of the “smooth HRF”-estimation commands:

On a side note, if you are looking for methods in AFNI, you can browse this list:

… and there are also a list of programs by topic (this has to be updated):

Please let us know (esp. Gang) if you have any questions.


1 Like

Hi Steven,

Thank you for your kind help. I will try the toolbox you provided. I will provide my solution if I have any results. Thank you for your attention again!


Hi @ptaylor,

Thank you for your detailed reply! The article you provide is exactly what I want, although I am not familiar with some terms mentioned in this article, such as different HRF models including “piecewise linear splines” or “cubic splines”. Anyway, I am trying to understand these new concepts or models first. Then, I may try the new AFNI tool “3dMSS” to analyze my task-fMRI data with “smooth splines”. If you have some information about the traditional TENT model or relating backgrounds, could you please share some with me to help understanding this knowledge? Thank you again!

Best regards,

Sure, about TENT modeling, here is an AFNI Academy video about it (might make sense to start at the beginning of this playlist, too, for background):
AFNI Time Series Analysis (part 4/8): FMRI Variable-Shape Models

There is a link in the text below the video to the relevant PDF and a script.

Re. cubic splines, that paper is introducing them, so a good description is there (TENT functions have been in AFNI for a long time). Essentially, they are a way to fit points to a smooth curve, which is then analyzed further.