Question
Question Posted 05/04/20:
Hello!
First, thanks for your time and for the database!
I have a couple of questions about the images found in the ADNI database.
First and foremost, I have been looking for raw DWI images, yet I have been unsuccessful to find them :(. Ideally, I would like to obtain raw DWI with >= 54 directions with multiple b values [1000,2000,3000*] (*preferably) and both directions for EPI correction, or in case, already pre-processed images. I observed that such images are described in the ADNI | Data Types page in the DTI section as: "Paul Thompson (USC) and Charles DeCarli (UCD) provide analyses of regional fractional anisotropy"; Yet I cannot find them in the database.
Have I been missing something? I have searched for diffusion images, but I can only access images that are termed “DTIâ€
So my second question is, what are exactly this DTI images ? Are they raw DWI? or are they pre-processed DWI, are they another type of image? Do they have a specific pre-processing pipeline?
I have been able to search for directions in the DTI, but not for the b-value, is there a form ?
Thanks for your time and have a nice day
Hello!
First, thanks for your time and for the database!
I have a couple of questions about the images found in the ADNI database.
First and foremost, I have been looking for raw DWI images, yet I have been unsuccessful to find them :(. Ideally, I would like to obtain raw DWI with >= 54 directions with multiple b values [1000,2000,3000*] (*preferably) and both directions for EPI correction, or in case, already pre-processed images. I observed that such images are described in the ADNI | Data Types page in the DTI section as: "Paul Thompson (USC) and Charles DeCarli (UCD) provide analyses of regional fractional anisotropy"; Yet I cannot find them in the database.
Have I been missing something? I have searched for diffusion images, but I can only access images that are termed “DTIâ€
So my second question is, what are exactly this DTI images ? Are they raw DWI? or are they pre-processed DWI, are they another type of image? Do they have a specific pre-processing pipeline?
I have been able to search for directions in the DTI, but not for the b-value, is there a form ?
Thanks for your time and have a nice day
Response posted 05/04/20 by Rob Reid:
Hello,
The raw images are available as DICOM. The processed images have been through a pipeline that (IIRC, it's not my pipeline) denoises, corrects for head motion and eddy current distortion, fits diffusion tensors, and outputs downloadable FA and MD images in NIFTI format. To work with the raw images you will need to implement a similar pipeline. If you start with dcm2niix it will output both a raw .nii and .bval and .bvec text files with the diffusion weightings in FSL format (also usable by dipy).
Unlike the Human Connectome Project, ADNI does not have any b=3000 images, or flipped phase encoding directions for correcting EPI distortion. We tried acquiring flipped phase encodings, but it turned out to be impractical in a study with multiple sites and scanner manufacturers. I recommend either using an analysis that does not need exact registration to an undistorted image, or literally nonlinearly registering to an undistorted image (the T1w).
ADNI has 3 main kinds of dMRI: ADNI2 had 5 b=0 + 41 b=1000 with a resolution of 2.7mm, from the GE scanners only. ADNI3 has dMRI from all of the scanners (GE, Philips, and Siemens) with 2.0 mm resolution. The Basic protocol has 6 b=0 + 48 b=1000 volumes. The Advanced protocol has those plus 6 b=500 + 60 b=2000 volumes.
The raw images are available as DICOM. The processed images have been through a pipeline that (IIRC, it's not my pipeline) denoises, corrects for head motion and eddy current distortion, fits diffusion tensors, and outputs downloadable FA and MD images in NIFTI format. To work with the raw images you will need to implement a similar pipeline. If you start with dcm2niix it will output both a raw .nii and .bval and .bvec text files with the diffusion weightings in FSL format (also usable by dipy).
Unlike the Human Connectome Project, ADNI does not have any b=3000 images, or flipped phase encoding directions for correcting EPI distortion. We tried acquiring flipped phase encodings, but it turned out to be impractical in a study with multiple sites and scanner manufacturers. I recommend either using an analysis that does not need exact registration to an undistorted image, or literally nonlinearly registering to an undistorted image (the T1w).
ADNI has 3 main kinds of dMRI: ADNI2 had 5 b=0 + 41 b=1000 with a resolution of 2.7mm, from the GE scanners only. ADNI3 has dMRI from all of the scanners (GE, Philips, and Siemens) with 2.0 mm resolution. The Basic protocol has 6 b=0 + 48 b=1000 volumes. The Advanced protocol has those plus 6 b=500 + 60 b=2000 volumes.
