Dear All,
I’m new to working with the ADNI dataset and I’d really appreciate any help or guidance.
I’m trying to download the T1-weighted MRI images from ADNI to use them with Clinica. However, I’m confused about how to do the image selection properly. I’ve read the instructions in the Clinica documentation:
, but if I follow them directly and download all images without any filtering criteria, the amount of data is enormous:
I also noticed that in one of the papers published by the authors of Clinica:
, they used a specific subset of ADNI data, but I don’t understand how they selected those images in practice. I’m not sure if this subset can be directly selected from the ADNI website, or if you have to download everything first and then select the relevant images afterwards.
How do researchers typically select a manageable and relevant subset of T1w MRI images from ADNI for studies using Clinica? Is there a standard approach or pipeline for this? Which filters should I select?:
I’d be extremely grateful for any advice, resources, or tips!
Thank you so much in advance for your help!
Best,
Elisa.
I guess Clinica selects a specific image ID for each subject and session based on clinical metadata. You can find details in the documentation here: https://aramislab.paris.inria.fr/clinica/docs/public/dev/Converters/ADNI2BIDS/#step-1-image-selection. This means that instead of downloading all available scans, you can identify the specific image per subject/session and download only those. In previous versions of Clinica, I was able to run the adni-to-bids converter using only the clinical data (without any image files). The output included a TSV file listing the selected image IDs (see screenshot below). Then, I simply used those image IDs to download the corresponding scans directly from the ADNI website rather than downloading the entire dataset.
However, in more recent versions of Clinica, this process no longer works as expected. The adni-to-bids converter now throws an error if no image files are present. One workaround is to install an earlier version of Clinica, e.g., I used version 0.8.4. There’s another issue: due to updates on the ADNI website, a clinical file called MPRAGEMETA.csv—which is necessary for image selection scripts—is no longer available. The converter fails without it. I’m not sure if there's any solution to this yet.
I am a new Clinica developer. I can try to help.
To my knowledge, we do not have a standard approach or pipeline for downloading the ADNI dataset other than what is stated in the Clinica documentation.
If you want to download the T1w modality data, you can check the appropriate box under the "IMAGING PROTOCOL" category in "Advanced Search".
Since ADNI is a large dataset, downloading all its data can indeed take several hours.
As Yuanwang Zhang already mentioned, you can select a specific list of subjects to make the download shorter, depending on your goals and the specifications of your study.
The LONI image and data archive has redesigned its ADNI data download interface, so the documentation and converter code may not be up to date, particularly regarding the downloading and processing of clinical data.
If you encounter any issues concerning the use of clinical data, please feel free to reach out again.
Best regards,
Adam ISMAILI