help
488 Subscribers 
help > Problem with second-level analysis main CONN GUI & result explorer display , when using different template and atlas files.
Showing 1-7 of 7 posts
May 23, 2019 04:05 PM | Mohammad Iyas Kawas - WFU HEALTH SCIENCES
Problem with second-level analysis main CONN GUI & result explorer display , when using different template and atlas files.
Hello Alfonso and CONN users,
I pre-processed and analyzed my data set in CONN, and I have used a different template and atlas (used proper TPM file with the correct corresponding atlas, I have also used corresponding brain mask for analysis in Setup>options>analysis mask).
I am now looking at my second level results and have changed the GUI background structural to my template T1 and changed the labels to the corresponding atlas file.
Now I see clear miss-labeling to some areas when I hover my mouse over the display.
Also, the results explorer is improperly labeling the clusters, and when I export the mask from the results explorer window, I get a clusters mask that has different dimensions than the T1 image. (results mask : 89,113,92) Vs (T1: 121,145,121 / same as corresponding atlas).
First, is the analysis affected by this problem and are the results and clusters still valid?
Second, can I use the voxel coordinates (x,y,z in mm) from the GUI and result explorer reliably to create seeds?
Third, How can I fix this or at least generate a proper results mast that I can overlay onto my atlas and calculate the number of cluster voxels in the corresponding ROI's.
Any Help is much appreciated,
Mohammad Kawas
I pre-processed and analyzed my data set in CONN, and I have used a different template and atlas (used proper TPM file with the correct corresponding atlas, I have also used corresponding brain mask for analysis in Setup>options>analysis mask).
I am now looking at my second level results and have changed the GUI background structural to my template T1 and changed the labels to the corresponding atlas file.
Now I see clear miss-labeling to some areas when I hover my mouse over the display.
Also, the results explorer is improperly labeling the clusters, and when I export the mask from the results explorer window, I get a clusters mask that has different dimensions than the T1 image. (results mask : 89,113,92) Vs (T1: 121,145,121 / same as corresponding atlas).
First, is the analysis affected by this problem and are the results and clusters still valid?
Second, can I use the voxel coordinates (x,y,z in mm) from the GUI and result explorer reliably to create seeds?
Third, How can I fix this or at least generate a proper results mast that I can overlay onto my atlas and calculate the number of cluster voxels in the corresponding ROI's.
Any Help is much appreciated,
Mohammad Kawas
May 23, 2019 06:05 PM | Mohammad Iyas Kawas - WFU HEALTH SCIENCES
RE: Problem with second-level analysis main CONN GUI & result explorer display , when using different template and atlas files.
An Additional note:
I have noticed that dimensions of the functional Nifti files after pre-processing are (91,109,91) but my analysis mask (used as explicit analysis mask in Setup>options) has the following dimensions (121,145,124 same as the template).
Could that be the issue ?? Does the explicit analysis mask needs to have certain dimensions? or match the (91,109,91) dimensions?
I hope this is just a visual error and not an issue that affects the statistical results in Conn when using different templates.
Thanks in advance,
Mohammad Kawas
I have noticed that dimensions of the functional Nifti files after pre-processing are (91,109,91) but my analysis mask (used as explicit analysis mask in Setup>options) has the following dimensions (121,145,124 same as the template).
Could that be the issue ?? Does the explicit analysis mask needs to have certain dimensions? or match the (91,109,91) dimensions?
I hope this is just a visual error and not an issue that affects the statistical results in Conn when using different templates.
Thanks in advance,
Mohammad Kawas
May 25, 2019 05:05 PM | Alfonso Nieto-Castanon - Boston University
RE: Problem with second-level analysis main CONN GUI & result explorer display , when using different template and atlas files.
Hi Mohammad,
Not sure about this yet. It is perfectly fine that the nifti files after pre-processing have (91,109,91) dimensions (that is typically the result of resampling after normalization), while your connectivity images in CONN have (121,145,124) dimensions (that is typically resampled to the size and voxel-resolution of your analysis mask as defined in Setup.Options), while the second-level masks generated by the results explorer window are somewhat smaller (the results explorer window is just cropping some out-of-brain areas for convenience). CONN handles all of these differences without a problem because all of those files are effectively co-registered (the nifti header voxel-to-world map contains the information about how to translate between those voxel dimensions and real-world spatial coordinates). So if you are seeing a mismatch between the reported voxel-coordinates there (in mm) and region labels (as defined by the background reference atlas in Tools.GUI options) that may indicate that some of those external sources (background image template and reference atlas) may not be in the same space / coregistered to your data. Could you please provide more details about your preprocessing steps (in particular what type of between-subject normalization or coregistration you performed) as well as about the source of the background anatomical image and reference atlas?
Thanks
Alfonso
Originally posted by Mohammad Iyas Kawas:
Not sure about this yet. It is perfectly fine that the nifti files after pre-processing have (91,109,91) dimensions (that is typically the result of resampling after normalization), while your connectivity images in CONN have (121,145,124) dimensions (that is typically resampled to the size and voxel-resolution of your analysis mask as defined in Setup.Options), while the second-level masks generated by the results explorer window are somewhat smaller (the results explorer window is just cropping some out-of-brain areas for convenience). CONN handles all of these differences without a problem because all of those files are effectively co-registered (the nifti header voxel-to-world map contains the information about how to translate between those voxel dimensions and real-world spatial coordinates). So if you are seeing a mismatch between the reported voxel-coordinates there (in mm) and region labels (as defined by the background reference atlas in Tools.GUI options) that may indicate that some of those external sources (background image template and reference atlas) may not be in the same space / coregistered to your data. Could you please provide more details about your preprocessing steps (in particular what type of between-subject normalization or coregistration you performed) as well as about the source of the background anatomical image and reference atlas?
Thanks
Alfonso
Originally posted by Mohammad Iyas Kawas:
An Additional note:
I have noticed that dimensions of the functional Nifti files after pre-processing are (91,109,91) but my analysis mask (used as explicit analysis mask in Setup>options) has the following dimensions (121,145,124 same as the template).
Could that be the issue ?? Does the explicit analysis mask needs to have certain dimensions? or match the (91,109,91) dimensions?
I hope this is just a visual error and not an issue that affects the statistical results in Conn when using different templates.
Thanks in advance,
Mohammad Kawas
I have noticed that dimensions of the functional Nifti files after pre-processing are (91,109,91) but my analysis mask (used as explicit analysis mask in Setup>options) has the following dimensions (121,145,124 same as the template).
Could that be the issue ?? Does the explicit analysis mask needs to have certain dimensions? or match the (91,109,91) dimensions?
I hope this is just a visual error and not an issue that affects the statistical results in Conn when using different templates.
Thanks in advance,
Mohammad Kawas
May 25, 2019 06:05 PM | Mohammad Iyas Kawas - WFU HEALTH SCIENCES
RE: Problem with second-level analysis main CONN GUI & result explorer display , when using different template and atlas files.
Thank you Alfonso,
I am using the Mayo clinic adult life span template and atlas. https://www.nitrc.org/forum/forum.php?forum_id=8992.
I use the default preprocessing pipeline and when I am prompted with the (Segment/normalize/resample settings) window I specify the the TPM file for the mayo clinic template.
And in setup>ROI I use the corresponding atlas too (an 122 ROI modified AAL atlas that is coregistered the template.)
In Setup>Options I also Use the (GM+WM) Brain mask created from and provided with the template as my explicit analysis mask.
And when I reach the 2nd level analysis display I specify the reference T1 for this template also and then I specify the Atlas too.
All the files (TPM, T1, Atlas, and mask are co-registered) and I download them as such.
NOTE: I did one thing that seems to have corrected the labeling issue in the GUI's display, which is I modified the .txt file associated with my atlas as it started the first line with a 0_unknown and then started listing the ROI's starting from the second line 1_precentral,....... (as you can see in the attached image). I changed the .txt file for the atlas to be very similar to the .txt file of the default conn atlas. (this fixed the display, but I didn't rerun the Setup or rerun the analysis as the for ROI to ROI analysis CONN seems to have generated the seeds correctly from the Atlas file. Do you think I need to rerun the Setup and analysis?)
Also, I understand from you that if the analysis Mask has different dimensions but is properly co-registered to the TPM file CONN will handle it correctly ?
Hope this is correct and that the analysis is not affected by these issues.
Many thanks,
Mohammad Kawas
I am using the Mayo clinic adult life span template and atlas. https://www.nitrc.org/forum/forum.php?forum_id=8992.
I use the default preprocessing pipeline and when I am prompted with the (Segment/normalize/resample settings) window I specify the the TPM file for the mayo clinic template.
And in setup>ROI I use the corresponding atlas too (an 122 ROI modified AAL atlas that is coregistered the template.)
In Setup>Options I also Use the (GM+WM) Brain mask created from and provided with the template as my explicit analysis mask.
And when I reach the 2nd level analysis display I specify the reference T1 for this template also and then I specify the Atlas too.
All the files (TPM, T1, Atlas, and mask are co-registered) and I download them as such.
NOTE: I did one thing that seems to have corrected the labeling issue in the GUI's display, which is I modified the .txt file associated with my atlas as it started the first line with a 0_unknown and then started listing the ROI's starting from the second line 1_precentral,....... (as you can see in the attached image). I changed the .txt file for the atlas to be very similar to the .txt file of the default conn atlas. (this fixed the display, but I didn't rerun the Setup or rerun the analysis as the for ROI to ROI analysis CONN seems to have generated the seeds correctly from the Atlas file. Do you think I need to rerun the Setup and analysis?)
Also, I understand from you that if the analysis Mask has different dimensions but is properly co-registered to the TPM file CONN will handle it correctly ?
Hope this is correct and that the analysis is not affected by these issues.
Many thanks,
Mohammad Kawas
May 25, 2019 08:05 PM | Alfonso Nieto-Castanon - Boston University
RE: Problem with second-level analysis main CONN GUI & result explorer display , when using different template and atlas files.
Hi Mohammad,
Regarding the modification to the .txt file associated with your atlas file, that looks exactly right. The format "ROI-number ROI-name/label" for your .txt file works just fine when used as an ROI atlas file in CONN (because that functionality is using REX, which supports a number of different label formats; see https://www.nitrc.org/forum/message.php?... for more details), but unfortunately you are right and the "background reference atlas" option in CONN's GUI currently only supports the simpler "ROI-name/label" format for your reference atlas files. Thank you for figuring out the issue here, and I will take a closer look to see if I can easily add support for other more complex background reference atlas formats into the next release.
Regarding whether it is necessary to re-run the Setup/Denoising steps after the update to the .txt file associated with your AAL atlas, in general you would need to do so, but in this particular case it is not necessary since CONN was already interpreting correctly the original .txt file.
And regarding the mask dimensionality, yes, that is perfectly fine as well, the analysis mask defined in the Setup.Options 'analysis-mask' field can have any arbitrary voxel dimensions / resolution, and as long as it is properly co-registered to (i.e. in the same space as) your fully-preprocessed functional data (which should be in this case, from your description of preprocessing steps and analysis mask file), CONN will handle that correctly.
Best
Alfonso
Originally posted by Mohammad Iyas Kawas:
Regarding the modification to the .txt file associated with your atlas file, that looks exactly right. The format "ROI-number ROI-name/label" for your .txt file works just fine when used as an ROI atlas file in CONN (because that functionality is using REX, which supports a number of different label formats; see https://www.nitrc.org/forum/message.php?... for more details), but unfortunately you are right and the "background reference atlas" option in CONN's GUI currently only supports the simpler "ROI-name/label" format for your reference atlas files. Thank you for figuring out the issue here, and I will take a closer look to see if I can easily add support for other more complex background reference atlas formats into the next release.
Regarding whether it is necessary to re-run the Setup/Denoising steps after the update to the .txt file associated with your AAL atlas, in general you would need to do so, but in this particular case it is not necessary since CONN was already interpreting correctly the original .txt file.
And regarding the mask dimensionality, yes, that is perfectly fine as well, the analysis mask defined in the Setup.Options 'analysis-mask' field can have any arbitrary voxel dimensions / resolution, and as long as it is properly co-registered to (i.e. in the same space as) your fully-preprocessed functional data (which should be in this case, from your description of preprocessing steps and analysis mask file), CONN will handle that correctly.
Best
Alfonso
Originally posted by Mohammad Iyas Kawas:
Thank you Alfonso,
I am using the Mayo clinic adult life span template and atlas. https://www.nitrc.org/forum/forum.php?forum_id=8992.
I use the default preprocessing pipeline and when I am prompted with the (Segment/normalize/resample settings) window I specify the the TPM file for the mayo clinic template.
And in setup>ROI I use the corresponding atlas too (an 122 ROI modified AAL atlas that is coregistered the template.)
In Setup>Options I also Use the (GM+WM) Brain mask created from and provided with the template as my explicit analysis mask.
And when I reach the 2nd level analysis display I specify the reference T1 for this template also and then I specify the Atlas too.
All the files (TPM, T1, Atlas, and mask are co-registered) and I download them as such.
NOTE: I did one thing that seems to have corrected the labeling issue in the GUI's display, which is I modified the .txt file associated with my atlas as it started the first line with a 0_unknown and then started listing the ROI's starting from the second line 1_precentral,....... (as you can see in the attached image). I changed the .txt file for the atlas to be very similar to the .txt file of the default conn atlas. (this fixed the display, but I didn't rerun the Setup or rerun the analysis as the for ROI to ROI analysis CONN seems to have generated the seeds correctly from the Atlas file. Do you think I need to rerun the Setup and analysis?)
Also, I understand from you that if the analysis Mask has different dimensions but is properly co-registered to the TPM file CONN will handle it correctly ?
Hope this is correct and that the analysis is not affected by these issues.
Many thanks,
Mohammad Kawas
I am using the Mayo clinic adult life span template and atlas. https://www.nitrc.org/forum/forum.php?forum_id=8992.
I use the default preprocessing pipeline and when I am prompted with the (Segment/normalize/resample settings) window I specify the the TPM file for the mayo clinic template.
And in setup>ROI I use the corresponding atlas too (an 122 ROI modified AAL atlas that is coregistered the template.)
In Setup>Options I also Use the (GM+WM) Brain mask created from and provided with the template as my explicit analysis mask.
And when I reach the 2nd level analysis display I specify the reference T1 for this template also and then I specify the Atlas too.
All the files (TPM, T1, Atlas, and mask are co-registered) and I download them as such.
NOTE: I did one thing that seems to have corrected the labeling issue in the GUI's display, which is I modified the .txt file associated with my atlas as it started the first line with a 0_unknown and then started listing the ROI's starting from the second line 1_precentral,....... (as you can see in the attached image). I changed the .txt file for the atlas to be very similar to the .txt file of the default conn atlas. (this fixed the display, but I didn't rerun the Setup or rerun the analysis as the for ROI to ROI analysis CONN seems to have generated the seeds correctly from the Atlas file. Do you think I need to rerun the Setup and analysis?)
Also, I understand from you that if the analysis Mask has different dimensions but is properly co-registered to the TPM file CONN will handle it correctly ?
Hope this is correct and that the analysis is not affected by these issues.
Many thanks,
Mohammad Kawas
May 29, 2019 11:05 PM | Mohammad Iyas Kawas - WFU HEALTH SCIENCES
RE: Problem with second-level analysis main CONN GUI & result explorer display , when using different template and atlas files.
Thank you so much Alfonso.
One other minor issue, My template, atlas, and mask are all co-registered and are in the same space. But i noticed that they have an LPI orientation While the outcome functional images are in the RPI orientation (which is the same for the default conn template and atlas). Does this affect the analysis in any way? Does conn properly handle the orientation? And in what orientation should the T1 and atlas be when I use them for the display of the second level results?
Hope this is handled by conn and that it won't be an issue.
Many thanks,
Mohammad Kawas.
One other minor issue, My template, atlas, and mask are all co-registered and are in the same space. But i noticed that they have an LPI orientation While the outcome functional images are in the RPI orientation (which is the same for the default conn template and atlas). Does this affect the analysis in any way? Does conn properly handle the orientation? And in what orientation should the T1 and atlas be when I use them for the display of the second level results?
Hope this is handled by conn and that it won't be an issue.
Many thanks,
Mohammad Kawas.
Jun 8, 2019 12:06 PM | heni
RE: Problem with second-level analysis main CONN GUI & result explorer display , when using different template and atlas files.
me?
Originally posted by Mohammad Iyas Kawas:
Originally posted by Mohammad Iyas Kawas:
Thank you Alfonso,
I am using the Mayo clinic adult life span template and atlas.
I use the default preprocessing pipeline and when I am prompted with the (Segment/normalize/resample settings) window I specify the the TPM file for the mayo clinic template.
And in setup>ROI I use the corresponding atlas too (an 122 ROI modified AAL atlas that is coregistered the template.)
In Setup>Options I also Use the (GM+WM) Brain mask created from and provided with the template as my explicit analysis mask.
And when I reach the 2nd level analysis display I specify the reference T1 for this template also and then I specify the Atlas too.
All the files (TPM, T1, Atlas, and mask are co-registered) and I download them as such.
NOTE: I did one thing that seems to have corrected the labeling issue in the GUI's display, which is I modified the .txt file associated with my atlas as it started the first line with a 0_unknown and then started listing the ROI's starting from the second line 1_precentral,....... (as you can see in the attached image). I changed the .txt file for the atlas to be very similar to the .txt file of the default conn atlas. (this fixed the display, but I didn't rerun the Setup or rerun the analysis as the for ROI to ROI analysis CONN seems to have generated the seeds correctly from the Atlas file. Do you think I need to rerun the Setup and analysis?)
Also, I understand from you that if the analysis Mask has different dimensions but is properly co-registered to the TPM file CONN will handle it correctly ?
Hope this is correct and that the analysis is not affected by these issues.
Many thanks,
Mohammad Kawas
I am using the Mayo clinic adult life span template and atlas.
I use the default preprocessing pipeline and when I am prompted with the (Segment/normalize/resample settings) window I specify the the TPM file for the mayo clinic template.
And in setup>ROI I use the corresponding atlas too (an 122 ROI modified AAL atlas that is coregistered the template.)
In Setup>Options I also Use the (GM+WM) Brain mask created from and provided with the template as my explicit analysis mask.
And when I reach the 2nd level analysis display I specify the reference T1 for this template also and then I specify the Atlas too.
All the files (TPM, T1, Atlas, and mask are co-registered) and I download them as such.
NOTE: I did one thing that seems to have corrected the labeling issue in the GUI's display, which is I modified the .txt file associated with my atlas as it started the first line with a 0_unknown and then started listing the ROI's starting from the second line 1_precentral,....... (as you can see in the attached image). I changed the .txt file for the atlas to be very similar to the .txt file of the default conn atlas. (this fixed the display, but I didn't rerun the Setup or rerun the analysis as the for ROI to ROI analysis CONN seems to have generated the seeds correctly from the Atlas file. Do you think I need to rerun the Setup and analysis?)
Also, I understand from you that if the analysis Mask has different dimensions but is properly co-registered to the TPM file CONN will handle it correctly ?
Hope this is correct and that the analysis is not affected by these issues.
Many thanks,
Mohammad Kawas
Hello Mohammad Kawas
I have the same problem. I am using the
pediatric template.
I used the default preprocessing pipeline and in
the (Segment/normalize/resample settings) window I specified the
the TPM.nii file for the pediatric template. I have noticed that
dimensions of the functional Nifti files after pre-processing are
(91,109,91) but the dimensions of my template are (197,233,189),
Also the dimensions of the warped structural images (wWM,wCSF,wGM)
are different with my template.
I didn't understand is it a problem or not? What
should I do?
Would you please advise me?
Thanks in advance,
Hengameh