MRI data

[Mathieu Dubois]

Hi everybody,

I'm a beginner with ADNI and therefore I'm not used to all the terms and protocols so I apologize for stupid questions.

The research institute where I work has stored some ADNI data (mainly T1 images and clinical scores) but in its own way. Therefore I would like to use ADNIMERGE to sort that. The first step is to be able for each subject to list all the examinations, the associated scans (original and repeat and which one is preferred) and clinical scores (MMSE, ADAS, etc.).

I think that the mri3met, mri3gometa list some attributes of the preferred image so I should probably merge them with other df but I can't find which one and worse I don't find the image ID or scan ID in those tables.

Can anyone help?

Mathieu

[Mike Donohue]

Hello Mathieu -

Perhaps LONISID in ADNIMERGE::mrinclusio might help. Or LONI's IDA_MR_Metadata_Listing (available from Download > Study Data > MR Image Acquisition.

Mike

[Mathieu Dubois]

Hi Mike,

Thanks for the answer. Could you tell me what exactly is in ADNIMERGE::mrinclusio? Is that the image that are actually processed in the study? In that case it's will help a lot.

Mathieu

[Mathieu Dubois]

Hi Mike,

I have merged the tables adnimerge and mrinclusio (with the default value for the by argument which is all the common column names i.e. "RID", "VISCODE", "COLPROT", "ORIGPROT" and "EXAMDATE"). This yields a 4238 x 110 dataframe. If I'm correct this gives the visit and diagnostic for each processed image.

By the way, merging with "RID", "VISCODE" and "EXAMDATE" only yields 4243 rows. I'm not sure to understand the difference.

[Mathieu Dubois]

Hi,

I have noticed that some subjects has several similar rows in mrinclusio. For instance, subject 1018 has 3 m06 (2 on 2007-06-14 (one at 1.5T, one at 3T) and one on 2007-07-16) visits and 2 m12 visits (same day: one at 1.5T and one at 3T). Is that normal? How to choose which one is correct?

[Danielle Harvey]

Hi Mathieu,

 

In ADNI-1 (RID<2000), a subset (about 25%) of the participants was scanned at both 1.5T and 3T.  If a scan didn't meet quality control standards, a person may have been rescanned for a particular visit (this might explain the 3 scans for 1018 at m06).  Do you know if the scans you are using are those that have been B0-corrected, etc. by the group at Mayo or if you are using the original scans?  You might find the MP-RAGE Metadata Listing (under Imaging, ALL MR Image Acquisition) helpful because it identifies the "preferred" scan (the SeriesID for the Orig/Proc=="Processed" scan) since Mayo only Gradwarped the preferred scan.  Also, do you know if you are using scans from individuals across all 3 phases of ADNI (so RIDs into the 5000s)?  If you are only using ADNI-1 participants (RID<2000), the MRI Core put together a predefined set of preferred scans for different series of visits (collections are available for download through the IDA and the listings are available on the data download page under Study Info, ALL Data & Databases (ADNI 1.5T MRI Standardized Lists and ADNI 3T MRI Standardized Lists))  We have not yet put together similar lists for ADNI-GO/2.

 

Danielle

---

From: adni...@googlegroups.com [adni...@googlegroups.com]
Sent: Friday, March 13, 2015 8:39 AM
To: adni...@googlegroups.com
Subject: [adni-data] Re: MRI data

Hi,

I have noticed that some subjects has several similar rows in mrinclusio. For instance, subject 1018 has 3 m06 (2 on 2007-06-14 (one at 1.5T, one at 3T) and one on 2007-07-16) visits and 2 m12 visits (same day: one at 1.5T and one at 3T). Is that normal? How to choose which one is correct?

--
You received this message because you are subscribed to the Google Groups "Alzheimer's Disease Neuroimaging Initiative (ADNI) Data" group.
To unsubscribe from this group and stop receiving emails from it, send an email to adni-data+...@googlegroups.com.
To post to this group, send email to adni...@googlegroups.com.
Visit this group at http://groups.google.com/group/adni-data.
For more options, visit https://groups.google.com/d/optout.

[Mathieu Dubois]

Dear Danielle,

Thanks for your response. If I understand correctly the "rescan" is systematic but is included in processing only if original scan doesn't pass QC. Is that correct?

I also wanted to ask what the column SITE in mrinclusio corresponds to? I thought it is the same than SITEID in adnimerge but it seems different.

[Mathieu Dubois]

After rethinking about the problem, I was wondering how can I find the visit (sc, bl, etc.) corresponding to a filename (for instance ADNI_002_S_0938_MR_MPR__GradWarp__B1_Correction_Br_20070219175537853_S19852_I40981)? If I'm correct it's possible just using the image ID (40981 in the example).

I forgot to mention that I use only ADNI1 subjects for now.

[Danielle Harvey]

Hi Mathieu,

My understanding is that if there was not a usable scan, an attempt was made to get a rescan (in ADNI-1, protocol was to obtain two scans at every scanning session, so that one was likely to be usable, but if neither was usable, a rescan session was scheduled).

What are you hoping to use the SITE or SITEID variable for? I'm not sure what the difference between the two variables is. My understanding is that the PTID (general format: XXX_S_XXXX) has site_S_RID, which seems to match the variable in ADNIMERGE (at least in the file I have downloaded from December), so that's the variable I'd be tempted to use...

Danielle

________________________________________
From: adni...@googlegroups.com [adni...@googlegroups.com]
Sent: Friday, March 13, 2015 11:08 AM
To: adni...@googlegroups.com
Subject: Re: [adni-data] Re: MRI data

Dear Danielle,

Thanks for your response. If I understand correctly the "rescan" is systematic but is included in processing only if original scan doesn't pass QC. Is that correct?

I also wanted to ask what the column SITE in mrinclusio corresponds to? I thought it is the same than SITEID in adnimerge but it seems different.

--
You received this message because you are subscribed to the Google Groups "Alzheimer's Disease Neuroimaging Initiative (ADNI) Data" group.

To unsubscribe from this group and stop receiving emails from it, send an email to adni-data+...@googlegroups.com<mailto:adni-data+...@googlegroups.com>.
To post to this group, send email to adni...@googlegroups.com<mailto:adni...@googlegroups.com>.

[Danielle Harvey]

Image ID, Series ID (19852 in example below) or scan date should help you identify the visit (though if you are using the GradWarp scans, as it appears you are based on the filename below, you will want to use the MPRAGEMETA file (the MP-RAGE Listing file I mentioned earlier), since ImageID changes when the scan is processed (Gradwarped, etc.)
________________________________________
From: adni...@googlegroups.com [adni...@googlegroups.com]
Sent: Friday, March 13, 2015 11:21 AM

To: adni...@googlegroups.com
Subject: Re: [adni-data] Re: MRI data

After rethinking about the problem, I was wondering how can I find the visit (sc, bl, etc.) corresponding to a filename (for instance ADNI_002_S_0938_MR_MPR__GradWarp__B1_Correction_Br_20070219175537853_S19852_I40981)? If I'm correct it's possible just using the image ID (40981 in the example).

I forgot to mention that I use only ADNI1 subjects for now.

--
You received this message because you are subscribed to the Google Groups "Alzheimer's Disease Neuroimaging Initiative (ADNI) Data" group.

To unsubscribe from this group and stop receiving emails from it, send an email to adni-data+...@googlegroups.com<mailto:adni-data+...@googlegroups.com>.
To post to this group, send email to adni...@googlegroups.com<mailto:adni...@googlegroups.com>.

[Mathieu Dubois]

Hi Danielle,

My understanding is that if there was not a usable scan, an attempt was made to get a rescan (in ADNI-1, protocol was to obtain two scans at every scanning session, so that one was likely to be usable, but if neither was usable, a rescan session was scheduled).

Thanks for the clarification.
 

What are you hoping to use the SITE or SITEID variable for? I'm not sure what the difference between the two variables is.  My understanding is that the PTID (general format: XXX_S_XXXX) has site_S_RID, which seems to match the variable in ADNIMERGE (at least in the file I have downloaded from December), so that's the variable I'd be tempted to use...

I was precisely expecting that those variables are equal but it appears they are not. adnimerge$SITEID is clearly the site ID found in PTID. Therefore I was wondering what mrinclusio$SITE is? I just fear that it may be a bug.

[Mathieu Dubois]

Le vendredi 13 mars 2015 19:49:37 UTC+1, Danielle Harvey a écrit :

Image ID, Series ID (19852 in example below) or scan date should help you identify the visit (though if you are using the GradWarp scans, as it appears you are based on the filename below, you will want to use the MPRAGEMETA file (the MP-RAGE Listing file I mentioned earlier), since ImageID changes when the scan is processed (Gradwarped, etc.)

Thanks again for this clarification. My request for data access has not been accepted yet therefore I can't download the file you mention. Do you know if the information is ADNIMERGE? The mriserial dataframe for instance?

[Mathieu Dubois]

Image ID, Series ID (19852 in example below) or scan date should help you identify the visit (though if you are using the GradWarp scans, as it appears you are based on the filename below, you will want to use the MPRAGEMETA file (the MP-RAGE Listing file I mentioned earlier), since ImageID changes when the scan is processed (Gradwarped, etc.)

Thanks again for this clarification. My request for data access has not been accepted yet therefore I can't download the file you mention. Do you know if the information is ADNIMERGE? The mriserial dataframe for instance?

 Or maybe the mrimppro df?

[Mathieu Dubois]

Hi,

Just to understand:
  - in adnimerge, subject 022_S_0014 has a bl visit on 2005-11-04 (EXAMDATE column)
  - in MPRAGEMETA, the closest scan date (ScanDate column) is 2005-09-29

If I understand correctly, the subject got a screening on 2005-09-04 and a baseline visit on 2005-11-04 but the screening scans were kept. Is that correct?

The problem is that, as the files I have are named according to the scan date, I can't find a corresponding visit in adnimerge.
However, according to diagram on page 17 of the ADNI Procedure manual, the maximum time between screening and baseline scans is: 3+14+28+1 = 59 days. So if, the time span between a EXAMDATE and a scan date is less than 59 days, it's safe to associate those scans to this visit.

[Mathieu Dubois]

Note that with the 59-days heuristic, subject 131_S_0123 has problems:
  - according to adnimerge, bl is on 2006-02-07, m06 on 2006-08-17 and m12 on 2007-02-01
  - there is a scan made on 2006-11-01 which is more than 59 days from m06

[Danielle Harvey]

In ADNI-1, the 1.5T MRI scans were obtained at the screening visit – if individuals met all entry criteria, they were then sent on to the baseline visit (where more clinical info was obtained, CSF sample obtained if they consented to CSF, and FDG or 3T MRI if they had been randomized to that arm) .  The 1.5T MRI scan at screening is considered the “baseline” scan for the 1.5T series.  Scans were not always obtained the same day as the clinical visit (EXAMDATE).

 

 

 

From: adni...@googlegroups.com [mailto:adni...@googlegroups.com]
Sent: Tuesday, March 17, 2015 11:09 AM
To: adni...@googlegroups.com
Subject: Re: [adni-data] Re: MRI data

 

Hi,

--

You received this message because you are subscribed to the Google Groups "Alzheimer's Disease Neuroimaging Initiative (ADNI) Data" group.

To unsubscribe from this group and stop receiving emails from it, send an email to adni-data+...@googlegroups.com.
To post to this group, send email to adni...@googlegroups.com.

[Danielle Harvey]

Do you have access to the MRIMETA or MPRAGEMETA file now? (the previous email made reference to MPRAGEMETA) – both of those list the scan date and corresponding visit (MRIMETA has VISCODE while MPRAGEMETA has Visit)

 

 

 

From: adni...@googlegroups.com [mailto:adni...@googlegroups.com]
Sent: Tuesday, March 17, 2015 11:18 AM
To: adni...@googlegroups.com
Subject: Re: [adni-data] Re: MRI data

 

Note that with the 59-days heuristic, subject 131_S_0123 has problems:

  - according to adnimerge, bl is on 2006-02-07, m06 on 2006-08-17 and m12 on 2007-02-01
  - there is a scan made on 2006-11-01 which is more than 59 days from m06

--

You received this message because you are subscribed to the Google Groups "Alzheimer's Disease Neuroimaging Initiative (ADNI) Data" group.

To unsubscribe from this group and stop receiving emails from it, send an email to adni-data+...@googlegroups.com.
To post to this group, send email to adni...@googlegroups.com.

[Mathieu Dubois]

Hi Danielle,

Thanks again for the help (and sorry for the delay). I have downloaded all the tabular data from ADNI on Monday. Combining information from MRIMETA.csv and MRI3META.csv, I can reconstruct the information for almost all subjects (we were using 1.5T and 3T images and I'm not sure that this has been correctly recorded in my research group).

More precisely, I have concatenated MRIMETA and MRI3META (they have the same columns) and join them with the information I have from the filenames according to RID and EXAMDATE.

The only exception I found is for the screening visit of subject 013_S_0240:
  - according to the files I have, the scans were performed on 2006-03-20
  - according to MRIMETA, the scans were performed on 2006-03-10
  - according to research on IDA this afternoon, the study date is on 2006-03-20 (same that in file names)

Given that the archive date is 2006-03-21 (and that more than 4000 files are correctly detected) I was wondering if this is a mistake in MRIMETA.

[Danielle J Harvey]

Hi Mathieu,

Yes, if the date matches both the scan file name and the date in the IDA, it is possible there is an error in the MRIMETA file.

Danielle

---

From: adni...@googlegroups.com <adni...@googlegroups.com> on behalf of 'Mathieu Dubois' via Alzheimer's Disease Neuroimaging Initiative (ADNI) Data <adni...@googlegroups.com>
Sent: Thursday, March 19, 2015 2:59 PM

To: adni...@googlegroups.com
Subject: Re: [adni-data] Re: MRI data

[Mathieu Dubois]

Hi Danielle,

I have re-checked the screening visit of 013_S_0240 and I can confirm it's different under IDA and in MRIMERGE (2006-03-20 and 2006-03-10 respectively). Could that be corrected?

By the way, why the VISOCDE column is not in IDA? It gives short and unified visit code (bl, m06, etc.) and every time I worked with MD this is what they want to know.

[Mathieu Dubois]

Is the Series ID available throught the "advanced search (beta)" tab of IDA? I can't find it. It would be very useful for my purpose.

[Mathieu Dubois]

Hi,

I have noticed several strange cases in the results of an IDA query (including original and pre-processed images matching with sequence matching "MPR*" and no other restriction):
  1) some subjects lack an original image for some visits; for instance subject 002_S_0295 has visits on 2006-04-18, 2006-11-02, 2007-05-25, 2009-05-22, 2011-06-02 and 2012-05-10 but I can find original image only for 2011-06-02 and 2012-05-10. Is that normal? There are many subjects in this case. Worse, subject 002_S_0295 has no original image at all.

  2) on the other hand, some subjects have only the original" image for some visits; for instance subject 003_S_0907 has only original images for visit 2009-10-27 (there some later visit). This case seems to happen often in the last visit. Maybe it's related to the fact that those images didn't pass the QC.

Last but not least, I have noticed that some images were acquired with 2.9T. Can I safely assume that these images are in the 3T group?

[Mathieu Dubois]

Looking back at subject 002_S_0295 in the web interface, I have noticed that the original image sequence is "MP-RAGE" (therefore not matching "MPR*"). Most other original images use "MPRAGE" without "-" (which match "MPR*") plus some information.

Is this an error or a normal variation (in which case I will re-run the search)?

[Danielle J Harvey]

Hi Mathieu,

 

If you look in the IDA MR Metadata Listing file (available for download on the Study Data page), there is a “Sequence” variable which gives you an idea of all of the many variations of sequence descriptors – MPR* will capture a lot of the MPRAGE, but not all (there are some that have MP-RAGE and MP RAGE, for example).  In addition to the variations on MPRAGE, IR-SPGR and IR-FSPGR are also 3D T1 weighted scans (by GE), so if you are looking for all of the 3D T1 scans, you need to also look for those scans.

 

Yes, I would assume the 2.9T is 3T.

 

Danielle

 

 

 

From: adni...@googlegroups.com [mailto:adni...@googlegroups.com]
Sent: Wednesday, March 25, 2015 3:53 AM
To: adni...@googlegroups.com
Subject: Re: [adni-data] Re: MRI data

 

Looking back at subject 002_S_0295 in the web interface, I have noticed that the original image sequence is "MP-RAGE" (therefore not matching "MPR*"). Most other original images use "MPRAGE" without "-" (which match "MPR*") plus some information.

Is this an error or a normal variation (in which case I will re-run the search)?

--

[Mathieu Dubois]

Hi Danielle,

Thanks for the answer. I have tried to use the Sequence field but it's a bit complicated.

Before your answer, I have also noted that some sequences are called "Sag MPRAGE" which I guess is MPRAGE. I have done another query for all images that match "*MP*" (it yields more than 60390 images). If I understand correctly, I should also take into account "IR-SPGR" and "IR-FSPGR".

Anyway I have made large progress (being able to match almost everything I need) by merging with MPRAGEMETA.csv file you recommended before (with respect to the Image ID). With that I have access to the Series ID (I have a doubt on one image but I will recheck first) and the field strength (information that I need) and may be able to reconstruct the thickness (needed too - could be in MPRAGEMETA).

Therefore I think I will make another query without any restriction and merge with MPRAGEMETA.csv.

Mathieu

[Mathieu Dubois]

Hi,

I have rechecked the image for which I had doubts in MPRAGEMETA. The image ID of the image is 149535. In MPRAGEMETA, this image is associated to series ID 65648. It the only image in this series.

I'm not sure that helps but there is an "original" image for the same subject and visit (image ID 140652) without other images in its series (series ID 65649). Given that the number are rather close could that be an error in MPRAGEMETA?

Can somebody confirm that? If so, where can I report such errors?

Mathieu

[Danielle J Harvey]

Hi Mathieu,

 

I would recommend sending a question through the ADNI “Ask the Experts” page (http://adni.loni.usc.edu/support/experts-knowledge-base/ask-experts/) – directed to the MRI Core, since they “process” the original images to produce the Gradwarped images.  They can probably look into this specific case and give you some information.

 

Danielle

 

From: adni...@googlegroups.com [mailto:adni...@googlegroups.com]
Sent: Thursday, March 26, 2015 8:16 AM
To: adni...@googlegroups.com
Subject: Re: [adni-data] Re: MRI data

 

Hi,

I have rechecked the image for which I had doubts in MPRAGEMETA. The image ID of the image is 149535. In MPRAGEMETA, this image is associated to series ID 65648. It the only image in this series.

Im' not sure that helps but there is an "original" image for the same subject and visit (image ID 140652) without other images in its series (series ID 65649). Given that the number are rather close could that be an error?

Can somebody confirm that? If so, where can I report such errors?

Mathieu

--

[Mathieu Dubois]

Thanks, I will do that ASAP.

[Mathieu Dubois]

Hi,

I have reported the error but haven't received an answer yet.

As we have seen above, the "description" column for original images is rather variable (MPRAGE, MP-RAGE, etc). However, I think that for processed images, it is more controlled. For example, if I'm correct, for gradwarp and B1 corrected images, the description always starts with "MPR; GradWarp; B1 Correction <-" for first scan and  "MPR-R; GradWarp; B1 Correction <-" for the repeated scan (the string behind the "<-" is the name of the original sequence).

Is it OK to assume that? This would give a way to identify the images.

Mathieu

[Mathieu Dubois]

Hi,

I have noticed that the expert contact form mention to not include subject identifiers. Does it means that my message has been blocked?

Mathieu

[Danielle J Harvey]

Hi Mathieu,

 

I am not sure – you can try emailing the Mayo group directly (adn...@mayo.edu)

 

Danielle

 

From: adni...@googlegroups.com [mailto:adni...@googlegroups.com]
Sent: Tuesday, April 07, 2015 8:49 AM
To: adni...@googlegroups.com
Subject: Re: [adni-data] Re: MRI data

 

Hi,

I have noticed that the expert contact form mention to not include subject identifiers. Does it means that my message has been blocked?

Mathieu

--

[Mathieu Dubois]

Thanks Danielle but the expert group answered a few minutes after this message.

[Mathieu Dubois]

Hi again,

The DATADICT.csv file states that the file MRIMPPRO.csv has a LONIUID column to identify the preferred image. I can't find it in the file I downloaded on March 16h. Is that normal? How can I identify the preferred image?

Mathieu

[Danielle J Harvey]

Hi Mathieu,

 

Generally speaking, Mayo only gradwarped, etc. the preferred image, so if you are using the gradwarped images, you should be getting the preferred scan (there may be some instances of multiple gradwarped, etc. scans, but for the most part, there will only be one for a given visit and field strength).  Since you are focusing on ADNI-1, you can also use the standardized lists of scans (under Study Info on the study data page) which provide lists of scans that the MRI core recommends be used in analyses (there are separate lists for bl only, bl+m06+m12, etc.).  Image collections have also been created in the IDA for these, so users can easily download the scans that are on those lists.

 

Danielle

 

From: adni...@googlegroups.com [mailto:adni...@googlegroups.com]
Sent: Wednesday, April 08, 2015 3:31 AM
To: adni...@googlegroups.com
Subject: Re: [adni-data] Re: MRI data

 

Hi again,

The DATADICT.csv file states that the file MRIMPPRO.csv has a LONIUID column to identify the preferred image. I can't find it in the file I downloaded on March 16h. Is that normal? How can I identify the preferred image?

Mathieu

--

[Mathieu Dubois]

Hi,

I have tried to use MPRAGEMETA. My idea was to find for each visit and field the series ID with the maximum number of images within.

Unfortunately, I have found a lot of cases where several series ID have the same number of images:
  - either the first scan and the repeat were processed
  - either a later scan (a rescan I guess) was processed too

I was thinking of using the MRIQUALITY.csv file but this file doesn't have the image ID and the LONIUID column (which according to the datadict is the series ID) does not match any series ID in my IDA CSV.

Similarly, the mrimppro table in ADNIMERGE has a columns named LONIUID but I can't find those values in my IDA CSV...

The standardized list of scans is not an option because my (current) goal is to extend the work of Cuignet et al. in "Automatic classification of patients with Alzheimer's disease from structural MRI: a comparison of ten methods using the ADNI database." (at the beginning of the discussion it was a different goal). Therefore I would like to all the gradwarp-b1 images for the 509 subjects in this study.

I'm sorry to ask so many questions but I'm a bit confused by all this.

Mathieu

[Danielle J Harvey]

Hi Mathieu,

 

Please send me an email directly (djha...@ucdavis.edu), so I can best help you figure out how to get what you want – this level of detail is probably not informative for the ADNI google group.

 

Danielle

 

From: adni...@googlegroups.com [mailto:adni...@googlegroups.com]
Sent: Thursday, April 09, 2015 9:58 AM
To: adni...@googlegroups.com
Subject: Re: [adni-data] Re: MRI data

 

Hi,

I have tried to use MPRAGEMETA. My idea was to find for each visit and field the series ID with the maximum number of images within.

Unfortunately, I have found a lot of cases where several series ID have the same number of images:
  - either the first scan and the repeat were processed
  - either a later scan (a rescan I guess) was processed too

I was thinking of using the MRIQUALITY.csv file but this file doesn't have the image ID and the LONIUID column (which according to the datadict is the series ID) does not match any series ID in my IDA CSV.

Similarly, the mrimppro table in ADNIMERGE has a columns named LONIUID but I can't find those values in my IDA CSV...

I'm sorry to ask so many questions but I'm a bit confused by all this.

Mathieu

--

[Mathieu Dubois]

Just to mention that I took a different strategy for this problem: I use MPRAGEMETA to identify original and fully pre-processed images which is what is inserted in my DB now.