Given two sets of rest state fmri images, one set has adhd and the other is control. The following t-test returns all 0’s for the p_values. Is this because the two images are rest state images ? How to find the difference between the images in such cases ?
from scipy import stats
smooth_img_class1=image.smooth_img(fmri_class_1.func_preproc[0], fwhm=3).get_data()
smooth_img_class2=image.smooth_img(fmri_class_2.func_preproc[0], fwhm=3).get_data()
_, p_values=stats.ttest_ind(smooth_img_class1, smooth_img_class2, axis=-1)
The p_values are all null. It is important to note that the smooth_img_class objectss are also all 0’s.
What is missing in this ?
Thanks for the help.
Are there NaNs in the input data before smoothing?
1 Like
I do not see nan. But after smoothing, the smooth_img_class1 and smooth_img_class2 arrays are all zeros! I tried changing the fwhm values to no avail.
Could you check your input data ?
I could not reproduce with ADHD datasets in nilearn (fetch_adhd)
1 Like
@KamalakerD Thank you. Sure. I think my question broadly is whether this example here: http://nilearn.github.io/auto_examples/04_manipulating_images/plot_roi_extraction.html#sphx-glr-auto-examples-04-manipulating-images-plot-roi-extraction-py can be applied to rest fmri (from control subjects and adhd subjects) also ?Or if there is another test (other than the t-test) to find the most intense voxels in rest state frmi.
The notion of ‘most instense voxels’ has little meaning in resting state. You need to use a linear decomposition method, as in the following example:
nilearn.github.io/auto_examples/03_connectivity/plot_extract_regions_dictlearning_maps.html
1 Like
orduek
March 20, 2019, 4:59pm
7
If I want to compare the changes in connectivity from time 1 to time 2?
1 Like