ICA extracts DMN of a single object (rsfMRI)

Hello, I also need to use ICA as a single object extraction network. But in the process of using it, I found some differences. An ADHD data and my data are shown below(Both use an object,one adhd data and one rsfMRI data). My data shows that there is something wrong with the results. Is the result normal? Is it my problem of data preprocessing? Or is there a problem with some codes? Is the extracted result correct?



num = 40
adhd_data = datasets.fetch_adhd(n_subjects=num)
func = adhd_data.func[0]
# func = r'I:\TenDrugDataNoBrain\Drug01_brain.nii'  # my data

canica = decomposition.CanICA(n_components=20, mask_strategy='background')
# # # func_filenames = rest_dataset.func
#Retrieving the components
components = canica.components_
# canica_components_img = canica.components_img_
#Using a masker to project into the 3D space
components_img = canica.masker_.inverse_transform(components)
canica_components_img = image.index_img(components_img, 9)
#Plotting the default mode network (DMN) without region extraction
plotting.plot_stat_map(image.index_img(components_img, 9), title='DMN')
# Plotting all the components
plotting.plot_prob_atlas(components_img, title='All ICA components')

You seem to have a lot of out-of-brain components. I suggest to do two things:

  1. Check with you preferred image viewer that your image is well aligned with the MNI template (e.g. Nilearn allows you to plot a functional volume overlayed with the outline of MNI template image)
  2. If the correspondence is OK, maybe you simply need to mask your iage better to keep only within-brain voxels. For this create a mask image and check it visually or use an MNI-derived brain mask.

Thank you for your suggestion, I will try it according to your method. In addition, I have a question, where can I get (eg: DMN, executive control network, working memory network, motor network) brain function network template, because with the template, I can realize what these networks look like.