Showing 1-13 of 13 posts
Display:
Results per page:
Jan 26, 2019  04:01 AM | Larry Olson - Emory University
SPECT
Are there plans to implement ISAS in BIS Web? Everything seems to be there except the statistical step with the normals.
Jan 28, 2019  11:01 AM | Xenophon Papademetris
SPECT
There is a test version hidden in plain sight:

https://bioimagesuiteweb.github.io/webap...

We would love some feedback on this.

Xenios
Jan 29, 2019  10:01 AM | Larry Olson - Emory University
RE: SPECT
BioImage Suite diff-SPECT- I have quickly tested this on several old images. Initial very preliminary impression is very good, but there are some confusing issues, and there are 2 things really important to me that I can't seem to address.

Good:
This is very promising.
I like that it is quick and produces results similar to legacy ISAS. User instructions will be important.
The Hyper- hypo- coordinate results are a really useful representation.
Great to be able to control overlay parameters, including in mosaic mode.
Text/csv result saving is great feature.
I would use this by default if I could save results.

Confusing:
Once registration is done, it shows the Object Map as an overlay. It is not obvious that user needs to select the Tmap as an overlay from the data tree. Perhaps this should default.
Clicking on a hyper-coordinate takes the cursor to a coordinate that is not really the most intense one, and does not appear to be the center of mass of the cluster.
Mosaic "labels" does not appear to show anything useful, just the slice boundaries.
I also wonder if the underlying method is identical to ISAS or to ISAS-HN? Or perhaps a modification similar to Mayo's Statiscom method? It was previously helpful to view the Subtraction results for reference. Now I am not entirely sure of what I am looking at.

*Main issue 1:
I think the displays are always in MNI space.
It is really important to me to have results or at least the "blobs" in patient space. I can't see a way to do this. Legacy ISAS allowed applying the Inverse transformation to the perfusion blobs, which could be exported to other programs for viewing and analysis on the actual patient image in a variety of ways, but
I don't see a way to display Tmap on patient image, and
I don't see a way to export the blobs in either patient or MNI space.

*Main issue 2: I will really look forward to having access to command line tools to accomplish this. I originally modified the SPM scripts for ISAS several years ago to simplify and more or less automate the process. This worked great. Unfortunately every Matlab and SPM update broke the process. Then BIS legacy implementation came along, and it was much easier to just use that than to try to keep up with SPM and Matlab. I hope that this new BIS tool will offer some command line control again. I would like to integrate the results into my other processing results.
Jan 29, 2019  11:01 AM | Xenophon Papademetris
RE: SPECT
Thanks for the feedback. We will get to this in the next few days and update the software. In general all you are asking for below is very doable but will take a little bit of time. The implementation is ISAS-HN.
Feb 15, 2019  11:02 AM | Larry Olson - Emory University
RE: SPECT
This web version of ISAS is quite fast, easy and nice. Thank you for allowing results to be saved.
2 requests:
-I am particularly interested in being able to export the TMap in the patient space. Legacy BIS allowed this. (Apply inverted transform to blobs, then save).
-Are there command-line tools to do this in an external pipeline. I am very keen on this.

Thanks for all your great work!
Feb 18, 2019  11:02 AM | Xenophon Papademetris
RE: SPECT
Larry,

Thanks. These are on the todo list. Hopefully early next month once the current set of deadlines is over.

XP
Mar 20, 2019  12:03 PM | Xenophon Papademetris
RE: SPECT
Larry,

Updated.


1. We can output native space now. Try https://bioimagesuiteweb.github.io/unsta...

After you Computed Diff Spect Maps, then click "Map TMAP to Native Space"
To see the images later go to Show diff SPECT Data Tree. Open the folder diff spect and and right click on either tmap or tmap native


1. There are updated command line tools (alpha!)

Download here: http://bisweb.yale.edu/binaries2/biswebn...
Use npm install to install these (as any node package, e.g. sudo npm install -g biswebnode-1.1.0-a9.tgz)

To run the diffspect

biswebnode diffspect -h

To save the tmap in native space you must specify both

--native true
--output2 somefilename.nii.gz

Here is an example:

biswebnode diffSpect -i Patient2ICTAL.nii.gz -r Patient2INTERICTAL.nii.gz --native true --nonlinear false --output ~/Desktop/diff.nii.gz --output2 ~/Desktop/diffnat.nii.gz

Xenios
Apr 2, 2019  06:04 PM | Larry Olson - Emory University
RE: SPECT
This SPECT analysis is generally pretty peppy. However,
After a very long wait, I was finally able to view the TMAP in "native space". However,
The TMAP is displayed over the interictal SPECT. This is very odd. I would think the preferred "native space" would be the T1 "patient MRI"

Moreover, the image is "frozen". I cannot move around in it.
I see no way the export the TMAP.

Finally, the other four SPECTs I tried never got this far even.

In addition, I wonder if it would be possible to export a TMAP that shows the features at a preferred p value?
Attachment: Capture.JPG
Apr 5, 2019  08:04 AM | Xenophon Papademetris
RE: SPECT
Larry,

Can we get one anonymized dataset and we will fix things?

XP
Apr 5, 2019  10:04 AM | Larry Olson - Emory University
RE: SPECT
Odd, shortly after I saw this ALL my BIS gmails disappeared. But yes. I attached the SPECTs and T1 for the case that did eventually have native space as the SPECT. The SPECTs come off PACs rotated, so they have been correctly reoriented (flipx, flipy), and T1 is skull stripped.

Thank you!
Attachment: SPECTs.zip
Apr 5, 2019  10:04 AM | Xenophon Papademetris
RE: SPECT
The attachments did not make this, please sent them directly to me at xenophon dot papademetris at yale dot edu.
Apr 17, 2019  01:04 PM | Larry Olson - Emory University
RE: SPECT
Originally posted by Larry Olson:
This SPECT analysis is generally pretty peppy. However,
After a very long wait, I was finally able to view the TMAP in "native space". However,
The TMAP is displayed over the interictal SPECT. This is very odd. I would think the preferred "native space" would be the T1 "patient MRI"

Moreover, the image is "frozen". I cannot move around in it.
I see no way the export the TMAP.

Finally, the other four SPECTs I tried never got this far even.

In addition, I wonder if it would be possible to export a TMAP that shows the features at a preferred p value?
FYI, I tried the command line tool for SPECT today. That is pretty easy to use, and quick, but the output is not coregistered to the MRI. These are the same images I previously zipped and forwarded to you.

I attached a snapshot of the xnrT1.nii.gz on the left yoked to the diffnat.nii.gz in the middle. On the right is the a simple linear registration of the two. The crosshairs in the MRIcron viewers should all be at the same anatomical location.

As far as I can tell the output is coregistered to the interictal SPECT, like in the Web version's output.
Apr 21, 2019  05:04 PM | Larry Olson - Emory University
RE: SPECT
I added the "alpha" command line tools For ISAS_HN to my Matlab script. This works really well.
-I specify the ictal and interictal files
-Reorient if needed
then they automatically:
-create the TMAP
-I do a linear registration to the T1 (since they are coregistered to the interictal at this stage now)
-Then I threshold the TMAP to intenisities >=2 and/or >=4 (sort of a arbitrary guess for now- see below)
-Then they overlay and display on the T1 or any file other in my pipeline. Outstanding.
No user input or supervision other than specifying the files.

The question I have for you is this:
If I remember correctly from BIS legacy, in the help menu there was a t-distribution "table" showing the critical points of the t-distribution interchangeably as T-values, Z-scores, or P-values. I am speculating here that there were 14 normal SPECT pairs, and to interchange them this would suggest df=13? Then a 2 tailed t-test p=0.001 would need T-values >=4.2, and p=0.01 would need T-values >=3.0, p=0.05 would need T-values >2.2?
Is this reasonable?

I would like to correct this point in my pipeline.
It adds an important additional analysis tool to the presurgical evaluations and makes it extremely easy for users.