Analysis of ASL images without M0 file

[ANINDITA BHATTACHARJEE Res. Scholar, School of Biomed. Engg. IIT (BHU)]

Hello,

I want to inform you that I am currently analyzing OASIS -ASL images for CBF calculation that don't have M0 files.

I can run the TestDataset of ExploreASL but unable to run the OASIS files.

I am new to this platform and unable to identify the source of errors.

I am attaching the google drive link for my sample files.

I will be highly appreciate it if you kindly help me to understand the problems for which the pipeline is not running.

https://drive.google.com/drive/folders/1xxS3rBg7qrS1rJrfvQ9lBxXLDVv5mN1_?usp=sharing

Regards,

Anindita Bhattacharjee

[Jan Petr]

Dear Anindita,

Thanks for your email and thanks for posting the whole folder - that makes answering your question much easier.

The issue is that ExploreASL expects an even number of volumes in the ASL4D.nii - full pairs of controls/labels - but there are actually 105 images instead.

It seems to me that the first image has a different contrast so it is probably an M0 image - but you have to verify this because I am not familiar with the OASIS data.

You thus need to modify the DataParameters_HiQ.json and add the following line:

"M0PositionInASL4D":[1],

This will remove the first volume from ASL4D.nii and put it to M0.nii.

If this is not really an M0 image (though it looks that it is), you can use this line instead:

"DummyScanPositionInASL4D":[1],

best,
Jna

[ANINDITA BHATTACHARJEE Res. Scholar, School of Biomed. Engg. IIT (BHU)]

Dear Jan,

Thank you for your valuable inputs.

We have tried to implement your suggestions both with and without M0 commands as in OASIS dictionary it is not mentioned that first image is M0 or not (though it seems to be based on the contrast).

We are getting more files but not the quantitative CBF files. I am again attaching the link for our current analysis.

https://drive.google.com/drive/folders/15aHYbbTdJ-M5O4F_975MM5vY7TkK240k?usp=sharing

I will be highly appreciated if you kindly inform me what more modification can be done to obtain the results.

Regards,

Anindita

--
You received this message because you are subscribed to the Google Groups "ExploreASL" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ExploreASL+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/ExploreASL/6c74dcef-55ab-4d08-bf82-fcc94220c8dan%40googlegroups.com.

[Jan Petr]

The issue is now with SliceTiming. 2D readouts need to have this value provided so that they can adjust the effective PLD for each slice based on its actual readout time. 

You have the SliceTiming value correctly provided in the ASL4D.json, however, it doesn't seem to be read correctly.

We had some issues with that, but they should all be fixed. I am not sure which version of ExploreASL are you using - but updating to the latest one should fix your problem.

The best is to use the version we are currently preparing for the 1.5.0 release. You can find it here:

https://github.com/ExploreASL/ExploreASL/tree/release-1.5.0

It will be released within few days until all tests are finished.

Let me know if you need any help with selecting the correct version and also if this helped or not.

best,
Jan

[ANINDITA BHATTACHARJEE Res. Scholar, School of Biomed. Engg. IIT (BHU)]

Dear Jan,

I want to inform you previously; we used ExploreASL version 1.4.0 for the analysis (test dataset were running).

As per your suggestion, we have used ExploreASL version 1.5.0, but it gives more errors, and the test data set provided in this version is also not working.

I am attaching the links for the TestDataset and my subject's analysis folders, which I have run in version 1.5.0.

https://drive.google.com/drive/folders/1iivrOtguE9Zg-XyPmMxSf6Bnz8U1yPlZ?usp=sharing

https://drive.google.com/drive/folders/1K-yo298_v8DywZaiW-yHnHXQALhagh06?usp=sharing

Thanks for your cooperation.

Regards,

Anindita

To view this discussion on the web visit https://groups.google.com/d/msgid/ExploreASL/8c18bbd0-5924-4bc8-985c-dcb57ae31400n%40googlegroups.com.

[Jan Petr]

Dear Anindita,

Thank you for your extensive testing - this is really helpful. 

The last result is rather odd. The FLAIR segmentation of WMH seems to fail. But I can't really reproduce this error myself.

Can you share some more details about your setup with me please:

Version of Matlab?

You are on Linux?

You've downloaded the ExploreASL from the GitHub web or you use directly the git client?

Was a clean version of previously unprocessed TestDataSet used?

Have you tried to exit Matlab before running the new version?

I am sorry for these trivial questions, but the LST toolbox for WMH segmentation of FLAIR images seems to run without any issues on my computer on both Linux and Windows. SPM tends to behave weird if something is changed in the code, so a complete restart of Matlab does help sometimes.

I will also try to 'finish' the processing in your data to see if there is something specific that doesn't work there.

best,

Jan

[Jan Petr]

Dear Anindita,

I have tried to process your data and it works - or at least it starts the LST segmentation of FLAIR. We might be facing some trivial issues with SPM - something like SPM being initialized with the previous version and having conflicts between ExploreASL 1.4 and 1.5.  Can you try to completely close all Matlabs. Start again in the ExploreASL_1.5.0, run ExploreASL_Initialize there and then proceed with analyzing your data - you can simply call ExploreASL_Master('...path/DataParms_HiQ.json') and it will continue where it left of and it should start the LST segmentation - see if it does it or immediately crashes.

Once we exclude these trivial issues, we can carry on with other tests, but I don't currently have a better idea what might be causing this.

best,

Jan

[ANINDITA BHATTACHARJEE Res. Scholar, School of Biomed. Engg. IIT (BHU)]

You are right about MATLAB. I have started freshly and able to run the test dataset of the 1.5.0 version.

Although my own data giving the same results that I have obtain with the previous version.

Are you able to run my sample data? If yes, then can you please share the folder with me.

Please find the folder of my current analysis.

https://drive.google.com/drive/folders/1Rmowlx_X12IYy4sNlXQ_KZ7Fuc-T9wMC?usp=sharing

Regards,

Anindita

To view this discussion on the web visit https://groups.google.com/d/msgid/ExploreASL/1da45cab-1c57-45a6-9671-f1db7ca4dc8fn%40googlegroups.com.

[Jan Petr]

Hi Anindita,

The issue is with the input parameters describing the ASL sequence. You can either specify them in the DataPar.json on study level. Then you want to use this format and add it to the DataPar.JSON:

"Q":{"BackgroundSuppressionNumberPulses":0,
                            "LabelingType":"PASL",
                            "InitialPostLabelDelay":2025,
                            "LabelingDuration":1650,
                            "SliceReadoutTime":46.7647}

Or, you can specify it on the subject level, but there, we have moved to the BIDS format so you have to enter it in s and not ms and use slightly different names:

"BackgroundSuppressionNumberPulses": 0,
"ArterialSpinLabelingType":"PASL",
"PostLabelingDelay":2.025,
"LabelingDuration":1.650

This is a bit confusing and we are working hard on making all this synchronized and all in the new ASL-BIDS format, but this takes time, so we have this hybrid version now. As a further comment: You actually don't need to provide the SliceReadoutTime as this is correctly obtained from the DICOM data.

Do provide either the extra info on subject or on study level, but preferably not on both - the subject info will be prefered, but it is better not to have info on two locations as this can lead to confusion. Advantages of study/subject level are obvious - less work on study level (edit on one place), more flexibility at subject level (but edit at more places).

To show what was the error - you have provided the data in the study-level format but in the subject-level file. For your convenience I have attached both updated files.

I hope that this explanation helps. Note that we also have routines for importing data from DICOM that can make these things easier.

Also - we can provide some guidance for the whole study. If you want, you can present your study and preliminary results on our ExploreASL user meetings that take place bi-weekly on Thursday. We would be happy to discuss the study design, QC, interpretation etc. If you are interested, simply let me know. From our experience - no offense meant - it is better to discuss potential issues before the processing is started, rather than to discuss potential issues in processing when the whole paper is written and there's much less motivation to change things :)

best,
Jan

[ANINDITA BHATTACHARJEE Res. Scholar, School of Biomed. Engg. IIT (BHU)]

Hello Jan,

Thanks a lot for your deep introspection.

With the JSON files you have provided and the clarification you have given, I can run my sample files.

However, I am getting negative and single-digit CBF values that are not correlated with standard ASL values.

I have also modified the values in the DataParameters_HiQ file but still getting the same results. (folder name 2.11-03-2021)

Therefore, I was wondering how to rectify these errors so that I can get the required values.

I am again sharing the analysis folder with you.

https://drive.google.com/drive/folders/1PwumH_PFjMlF0BXU1hwWSrn3tKqAnZVc?usp=sharing

https://drive.google.com/drive/folders/1-qbC3UjIC5MXyABoaLzYb48whHkKfkQn?usp=sharing

Thank you once again for your precious time and consideration.

Regards,

Anindita

[Jan Petr]

Hi Anindita,

Just very quickly... I have noticed that the values in DataPar_HiQ.json are in s instead of ms. We're working on conversion to full BIDS format so the individual JSONs in ASL4D.json are now in second, and the common study parameters are in milliseconds. Can you Fix it in the DataPar_HiQ.json - all the three numeric values - and re-run? Best to re-run from a scratch to ensure that no precalculated values are used.

There's a small chance that this will help, if it doesn't help and the values in the CBF.nii are still around 1-5, then let me know and I'll try to look more closer.

best,

Jan

[ANINDITA BHATTACHARJEE Res. Scholar, School of Biomed. Engg. IIT (BHU)]

---------- Forwarded message ---------
From: ANINDITA BHATTACHARJEE Res. Scholar, School of Biomed. Engg. IIT (BHU) <aninditabhattac...@itbhu.ac.in>
Date: Sun, Mar 21, 2021 at 3:12 PM
Subject: Re: Analysis of ASL images without M0 file
To: Jan Petr <jan.hon...@gmail.com>

Dear Jan,

I have implemented your suggestions and currently getting double digit CBF values in GM (lower values) and negative values of WM CBF (first drive link).

As per your suggestion I have tried with new subject but getting same kind of results (second drive link).

https://drive.google.com/drive/folders/1AvjMjzgD2ilTDaF0vkcymF0pjbZAogR4?usp=sharing

https://drive.google.com/drive/folders/1GcHH4DzsbKE0EVUKOFyOkddtP5ojzVqh?usp=sharing

Regarding your question related to PASL product, I have attached screenshot of OASIS data dictionary as well as the pdf document.

I am assuming there is some minor mistake in somewhere as I can see all the steps of ExploreASL (mentioned in the paper) are running properly.

Thank you for your kind co-operation and support.

Regards,

Anindita

On Fri, Mar 19, 2021 at 3:52 PM Jan Petr <jan.hon...@gmail.com> wrote:

HI Anindia,

Yes, you are right, the second folder was correct. And that wasn't the problem anyway, as in ASL4D.json, the values were saved in correct units and taken from there.

What actually was the problem is that Siemens uses a scaling of M0 images for certain sequences. I think that a default scaling of 10 is used for the PASL product you are using. But it seems that in your case a scaling of 1 was applied. The reason is either that the scaling is not in your sequence (do you have the PASL product?), or somebody modified it on the console, or that ExploreASL didn't recognize the sequence correctly. 

To fix this, simply add the M0_GMScaleFactor to the DataPar.json:

"M0_GMScaleFactor":0.1,

"Q":{"BackgroundSuppressionNumberPulses":0,
                            "LabelingType":"PASL",
                            "InitialPostLabelDelay":2025,
                            "LabelingDuration":1650,
                            "SliceReadoutTime":46.7647}

We are currently working on a better sequence recognition and fixing it automatically, but this parameter is a fine work-around for you.

Go to "lock/xASL_module_ASL/Sub-001/xASL_module_ASL_ASL_1" directory and delete all lock files from 060 up. And then re-run ExploreASL. The values should be now normal.

Regarding the data itself. The CBF seems a bit lower nevertheless and the quality is not the best either. But you might need to check how it looks for more subjects. Unfortunately, the old Siemens PASL without background suppression isn't the most performing sequence and it is rather susceptible to patient movement - which seems to have degraded the quality of this particular scan. But lets try to process more data - and we can discuss those on our user meeting.

best,
Jan

On Tue, Mar 16, 2021 at 10:00 AM ANINDITA BHATTACHARJEE Res. Scholar, School of Biomed. Engg. IIT (BHU) <aninditabhattac...@itbhu.ac.in> wrote:

Kindly check the second folder, I think I have change the values.