fNIRS 3D reconstruction

Discussions
,
1

Dear Nirstorm community,

we are currently working on a NIRS dataset on natural speech perception, and want to use the 3D reconstruction function to map our channel data on the brain surface. We encountered several issues, and would be interested in your thoughts/comments.

First of all, some technical details: We use a NIRx Nirsport 2 sytem with 16x16 optodes, as well as 8 short-channels. See below:

grafik
grafik1204×877 105 KB

Data was preprocessed in Homer, and we exported the cleaned and filterered data as SNIRF file to brainstorm - that worked fine :slight_smile:

We got the 3D reconstruction working in an example subject, but still have several questions on the overall procedure:

  1. We identified bad channels in each dataset using qtnirs in our preprocessing script. Should we just mark these channels (as well as the short-channels perhaps) as bad channels in brainstorm to exclude them from the 3D reconstruction? Or should we remove them from our channel list entirely?

  2. When computing the fluences model, we found that our wavelengths (760 nm and 850 nm) are not specified in the tissue_property.json file. For now, we just manipulated the entries to fit our wavelengths (as recommended here: Fluences for 762nm (Artinis Brite fNIRS)). However, it would be great to use the real values. Is there any way to compute those? Or can you perhaps update your json file to work with our NIRx system?

  3. The default setting for the fluences model is 10 million photons. Is this a good setting or should adjust it for real data analysis?

  4. For the 3D reconstruction, we provided brainstorm with the Differential Optical Density (dod) data from Homer3. The dod data is then converted automatically to hemoglobin concentrations during the source reconstruction (we used wMNE for now). However, we cannot set DPF method or subject age here, which we normally do. What method/settings are used for this conversion?
    Is there a way to apply the source reconstruction on HbO/HbR time courses directly (imported from Homer3)?

Here is an example output of our wMNE model:

grafik
grafik885×648 86.5 KB

Thanks a lot for your help!

Best wishes from Germany!
Sebastian

2

Ping @edelaire

1 Like
3

Hello Sebastian, I recommend looking at the NIRSTORM plugin and its tutorials, it might be helpful. I am not sure how to help with 2, but I hope this can help:

  1. Marking your bad channels should be sufficient. Once you've performed short channel regression, you can also mark the short channels as bad channels.
  2. 10 million photons is enough to get a good estimate of where your activity is occuring, but for more detailed activity you may want to use 100 million instead.
  3. NIRSTORM's dOD to [HbO] and [HbR] allows you to pick from 2 different DPF methods (Scholkmann, 2013 or Duncan 1996), which can then be used for cortex reconstruction. [https://neuroimage.usc.edu/brainstorm/Tutorials/NIRSTORM](The tutorial for reconstruction can be found here.)
1 Like