help
help > RE: error during Step 5/7: Importing ROI Data
Apr 11, 2020 11:04 PM | Alfonso Nieto-Castanon - Boston University
RE: error during Step 5/7: Importing ROI Data
Hi Marianna,
The error indicates that the number of voxels with >50% posterior probability values identifying them as CSF tissue is unusually low (typically you would expect to find around 300,000 CSF voxels before erosion, and around 70,000 after erosion, while in subject 16 there are only ~3,000 CSF voxels before erosion, and none after erosion). If the erosion mask looks nevertheless fine, then this simply means that the posterior probability values are generally low for this subject, indicating that SPM was not able to gain enough confidence on the normalization results to properly match this subject CSF tissue to the standard MNI reference-space (i.e. normalization did not quite work for this subject). If this subject has a significant brain lesion, perhaps one strategy to try to improve the quality of the normalization results would be to include during normalization a mask identifying the location of the lesion (see for example the following thread https://www.nitrc.org/forum/message.php?msg_id=28708 for details on how to do that)
Best
Alfonso
Originally posted by Marianna La Rocca:
The error indicates that the number of voxels with >50% posterior probability values identifying them as CSF tissue is unusually low (typically you would expect to find around 300,000 CSF voxels before erosion, and around 70,000 after erosion, while in subject 16 there are only ~3,000 CSF voxels before erosion, and none after erosion). If the erosion mask looks nevertheless fine, then this simply means that the posterior probability values are generally low for this subject, indicating that SPM was not able to gain enough confidence on the normalization results to properly match this subject CSF tissue to the standard MNI reference-space (i.e. normalization did not quite work for this subject). If this subject has a significant brain lesion, perhaps one strategy to try to improve the quality of the normalization results would be to include during normalization a mask identifying the location of the lesion (see for example the following thread https://www.nitrc.org/forum/message.php?msg_id=28708 for details on how to do that)
Best
Alfonso
Originally posted by Marianna La Rocca:
As you can see form the attached screenshot, I
don't see particular problems in the normalization and in the CSF
segmentation.
I have also tried to perform again all the preprocessing but I have the same error for the same subjects.
Many of the subjects that I am studying have big brain lesions. Could the problems be due to that?
Any other suggestion on how to fix the issue?
Thank you,
Marianna
I have also tried to perform again all the preprocessing but I have the same error for the same subjects.
Many of the subjects that I am studying have big brain lesions. Could the problems be due to that?
Any other suggestion on how to fix the issue?
Thank you,
Marianna
Threaded View
| Title | Author | Date |
|---|---|---|
| Marianna La Rocca | Apr 7, 2020 | |
| Marianna La Rocca | Apr 11, 2020 | |
| Alfonso Nieto-Castanon | Apr 11, 2020 | |
| Marianna La Rocca | Apr 13, 2020 | |
| Pravesh Parekh | Apr 7, 2020 | |
| Marianna La Rocca | Apr 7, 2020 | |
| Marianna La Rocca | Apr 7, 2020 | |
| Marianna La Rocca | Apr 7, 2020 | |
| Pravesh Parekh | Apr 8, 2020 | |