number of voxels in VC doesn't match in Shen2019 and Kamitani2017

[zjc062]

Hello,

Thank you so much for publishing these amazing data/toolbox/results.

I tried to get the data from both the Generic_Object_Decoding project (Kamitani2017) and the Deep Image Reconstruction project (Shen2019) with the following download link:
Kamitani2017: https://figshare.com/articles/dataset/Generic_Object_Decoding/7387130
Shen2019: https://figshare.com/articles/dataset/Deep_Image_Reconstruction/7033577

However, when I try to extract the voxels from visual cortex (i.e., 'ROI_VC'), I found these two datasets differ a lot in size.
The number of VC voxels in the Kamitani2017 dataset is 3444, 4979, 5355, 4656, 6237 for each subject respectivel;
however, the number of VC voxels in Shen2019 dataset is 11726 for all the participants.

I used bdpy toolbox to extract the voxels. And I used this function to get the VC ROIs - bdata.select('ROI_VC')

Could you please kindly have a look and see if this is normal?

Thanks!

[mitsuaki]

In both projects, we used different MRI acquisitions.
In the Generic_Object_Decoding project, we acquired EPI data with voxel size 3x3x3 mm, but with voxel size 2x2x2mm in the Deep Image Reconstruction project.
(So, the number of voxels in the Deep Image Reconstruction is much larger than that of the Generic Object Decoding, because we acquired with the same Field of View (FOV) for both projects. See also the "MRI acquisition" section in both papers)

however, the number of VC voxels in Shen2019 dataset is 11726 for all the participants.

As far as I have checked the data uploaded on Figshare, the number of VC voxels differs for each subject.
(e.g., the number of voxels for subject1, 2, and 3 are 11726, 11114, and 9919, respectively)
Please check the subject index at the beginning of the file name ('sub-01-***.h5), and if you still have the problem, let us know the specific file name.

Thanks.