Investigators wishing to use ADNI data must apply for access and obtain approval from the ADNI DPC. Once approval is received, users may log into the Image and Data Archive to select and download data. Additional information about using the LONI IDA may be found on the IDA Support page
The clinical data set may be downloaded by authorized users only. Please visit the Access Data page for more information. Note that not all clinical data files contain date of the exam (EXAMDATE) especially for ADNIGO/2. You can extract EXAMDATE from the registry table (REGISTRY.csv) using RID, Phase, and VISCODE.
It is located on the LONI Image Data Archive (IDA) . Enter your username and password, go to Download, then Study Data. When you click on Study Info, you will see Data Dictionary [ADNI1,GO,2] (DATADIC.csv). Some files will have their own data dictionary, particularly those related to some of the biospecimens and the imaging numeric summaries.
There are three versions of the ADNI cognitive assessments for ADNI. Unfortunately, we are unable to release the worksheets. Refer to Cognitive Assessment section of the ADNI General Procedures Manual for information regarding cognitive assessments at visits other than BL, M6 and M12.
You may find the images in the following ways:
Not currently.
Missing data is coded with -1 and -4. Generally, -4 is used for “passively” missing or not applicable (e.g. item is not collected at a particular visit), and -1 is used for data which is confirmed missing at point of data entry.
The study design specifies the visits required for each research group (Control, eMCI, MCI, AD). You can review the study schedule in the procedures manual for ADNI1, ADNI GO, or ADNI2 or download the study schedule. These documents can be found on the Documents: Procedures, Protocols, and Manuals webpage.
FOR SCREENING/BASELINE SCANS: A chart detailing imaging visits is shown in the Image Data section. Further explanation could be — (ADNI 1): Subjects who were randomized into the 3T scanner group had a 1.5T screening scan as part of the study inclusion/exclusion screening. Those subjects meeting inclusion criteria had a 3T scan at the Baseline visit.
FOR SCREENING/BASELINE DIAGNOSIS: The screening visit consists of a subset of the planned assessments (neuropsychological testing, imaging, etc.) sufficient to evaluate the inclusion/exclusion criteria for each potential participant. Once a participant passed screening, the baseline assessment was conducted in which additional neuropsychological testing, imaging, and fluid samples were taken. The diagnosis at the screening diagnosis was used to randomize individuals to an arm of the ADNI1 study (1.5T only, 1.5T + 3T, 1.5T + FDG-PET), but the diagnosis from the baseline visit is considered more accurate. For most individuals, the two diagnoses are the same, but the diagnosis did change for a handful of subjects.
You can extract the baseline diagnosis from LONI Download Study Data >> Diagnosis >> Diagnosis Summary [ADNI1, GO, 2] (DXSUM_PDXCONV_ADNIALL.csv). ADNI1 baseline diagnosis can be obtained using the “DXCURREN” variable at VISCODE==”bl,” ADNIGO/2 baseline diagnoses can be obtained using the “DXCHANGE” variable, and ADNI3 baseline diagnosis can be obtained using the “DIAGNOSIS” variable.
FreeSurfer can measure volume of cortical or subcortical structures and compute cortical thicknesses using cross-sectional and longitudinal processing. Longitudinal processing creates within a subject template and initializes each time point with the template to reduce individual variability. The ADNI1 images have been processed both cross-sectionally and longitudinally up to time point 9. ADNIGO/ADNI2 images have been processed using cross-sectional FreeSurfer and are being processed using Longitudinal FreeSurfer for subjects with at least 2 time points within 1 year. ADNIGO/2 longitudinal data will be processed to track longer-term changes as data accumulates. If the researcher is interested in baseline values and their association with some longitudinal or cross-sectional outcome, we recommend using cross-sectional FreeSurfer values. If interest lies in longitudinal change in the volumes or cortical thicknesses, the longitudinal version of FreeSurfer is recommended (when restricting to ADNI1 subjects).
Specimens from the same subject at the same time point will have slightly different test values if they are tested at different times. To minimize differences due to different assays, when a new batch of samples is run from later visits, samples from earlier visits are also included. Therefore, each biomarker file contains values from all samples run in the same batch and researchers should only use values within the same file.
Download the Cross-Validation training set from the Download | Study Data section of the ADNI data repository (https://ida.loni.usc.edu).
Using “REGISTRY.csv,” we have 822 subjects with initial screening (VISCODE == ‘sc’), but only 819 subjects had baseline visit conducted (VISCODE == ‘bl’ and RGCONDCT == 1 (RGCONDCT: Was this visit conducted? 1=Yes, 0=No)). Three people were deemed ineligible after the screening visit and were therefore not continued to the baseline visit. The 819 that had a baseline visit are considered the official ADNI1 sample.
The NV (“No Visit” or “Not Yet Determined”) visit code is applied when we receive scan QC data (e.g., mrinclusio) before we receive scan information sheets (e.g., mrimeta). In these instances, we do not yet know with which visit the scan QC data is associated.
Occasionally scan QC and scan information records which should be linked are not linkable because of discrepant data (e.g., exam dates do not match). The NV data code will be used in these cases as well, until the discrepancies are corrected.
The Alzheimer’s Disease Assessment Scale was devised to evaluate cognitive impairment in the assessment of Alzheimer’s disease. ADAS-Cog is recommended for second stage or more detailed assessments and/or for particular research evaluations rather than for applications in routine care settings.
In the ADNI subject IDs of the form 123_S_5678, the last 4 digits correspond to the roster ID (RID) which you see in every CSV file. You can also refer to the ROSTER table which lists RIDs and their associated PTIDs (of the form 123_S_5678).
Diagnosis conversion information is captured in the “Diagnostic Summary” data set which may be downloaded from within the ADNI Data Archive
ADCS created a new visit code variable called “VISCODE2” (sc, scmri, bl, m06, etc.) for ADNI2. (Please see ADNI2 Visit Codes Assignment Methods (PDF) on LONI). For longitudinal models, it is still recommended that you use the actual time since the initial visit (using EXAMDATE) rather than the visit code to determine time as there is variability in when the visits occurred.
Early MCI (eMCI) subjects were added to the recruitment in ADNIGO and ADNI2 to get some MCI subjects earlier during their impairment. These individuals still have a diagnosis of MCI but have less memory impairment than the MCI subjects recruited during ADNI1 (Late MCI or lMCI). ADNIGO and ADNI2 will continue to recruit subjects in the lMCI category. You can identify the eMCI subjects by merging the ARM table with the Diagnostic Summary table and identifying those with ARM==10 & DXCHANGE==2.
Part of the ADNI Data Use Agreement requires that you update your proposed analysis on an annual basis in order to keep your account activated. We send an email reminder a few weeks prior. If you fail to comply, your account will be deactivated. You may need to reactivate your account or reapply for data access. Contact the DPC to re-evaluate your account access.
Click here to recover your password and access your Data Archive Account.
The basic requirements are:
It takes approximately 2 week for your application to be approved. The online application form must be fully completed. In case you made an error, you can resubmit your application for ADNI Data. Specific inquiries can be directed to the ADNI Data Sharing and Publications Committee (DPC).
Although the four studies all aim to advance AD research, you will find differences in participant pools, and adjustments to data collection protocols over time as a result of scientific findings and changing technologies. These differences are outlined in the Study Objectives , and the Data & Samples sections for each core. For instance, DTI data is available from some ADNI GO and ADNI 2 subjects, as this is fairly new technology. ADNI manuals, procedures are protocols from each study are available on the ADNI Documents: Procedures, Protocols, and Manuals webpage for download. Visit the ADNI 3 page to learn more about what’s new in ADNI 3.
Acknowledgement requirements are detailed in the Data Sharing and Publications Policy. This and other relevant documents can be found on the webpage, “How to Apply for Data“.
Please visit the Alzheimer’s Disease Education and Referral (ADEAR) Center Website for a listing of open clinical studies. You can sign up at one of our acquisition site locations, in the US or Canada. Browse our map on the Study Site page.
Michael W. Weiner, MD is the principle investigator of ADNI. Read his bio and get his contact information from the Administrative Core webpage.
MR vendors now apply their own pre-processing steps as part of the product, prompting ADNI to not perform its own pre-processing on images for ADNIs 3 and 4. ADNIs 1, and the initial parts of ADNI2/GO images were captured prior to vendor-initiated pre-processing, so the MRI Core provided intensity normalized and gradient un-warped T1 image volumes during these ADNIs. Later in ADNI2/GO, these corrections were applied on the scanner at the time of image reconstruction when these had been implemented in the vendor product.
For ADNI4 data, it is recommended to use the “chosen” series selected during the QC process. Every series in every scan undergoes quality control at the Mayo ADIR Lab which includes selecting the “best” series if duplicate acquisition of that series was uploaded from the site. If the interest is in previous ADNIs, researchers may find a benefit from selecting the MPRAGE files for the highest quality ratings. Furthermore, imaging files are available at every stage of pre- and post-processing where applicable, and researchers are encouraged to utilize the detailed search functions when searching for and downloading images to suit their particular needs.
Paper protocols for all phases of ADNI can be found here
Although efforts are made to create similar naming conventions across vendors there is an innate difference between certain sequence names from different vendors. For example, a 3D T1 scan may be called an MPRAGE for Siemens and Philips, however, until recently, GE’s version of T1 was named IR-SPGR. Users should not rely solely on series naming conventions but look at the information located in the images DICOM header to determine exact parameters.
No, field mapping is not acquired in any ADNI4 scanner manufacturers. No other additional field maps are created/offered by ADNI our recommendation would be to use the Vendor products for field mapping/corrections.
The ADNI4 protocol includes a localizer, 3D T1, 3D FLAIR, 3D T2, T2 GRE, DTI, ASL, rsfMRI, and a High-Resolution Hippocampus scan. Spatial resolution for the 3D T1, 3D FLAIR, and 3D T2 series are all 1mm cubed. Please see ( ) for a more detailed description of the scans included in this protocol. A considerable effort was made to ensure harmonization across all available ADNI4 data regardless of scanner type, participant, and timepoints. However, the ADNI4 protocol has unique characteristics included compared to previous ADNI protocols, so care should be given when comparing data across ADNI iterations.
This information can most reliably be found in the DICOM headers of the fMRI acquisition. In addition, some of this information can be found in the NFQ data analysis provided for users located here: IDA
No, the fMRI performed in ADNI is intended to be a task-free resting state fMRI. Participants are only instructed to keep their eyes open during the fMRI sequence to ensure they are not sleeping during the acquisition and to achieve as uniform a functional state as possible across scans.
As with all sequences acquired, detailed information can most reliably be found in the DICOM headers of the acquisition. The protocol descriptions ( ) provide general parameter information for each series type by vendor platform. Information about diffusion analysis methods, including DTI, can be found at LONI: IDA
ADNI recommends that the main diffusion sequence be acquired with a P->A acquisition, unless you have a particular interest in the brainstem, in which case you should use A->P For many scanners ADNI also acquires a short A->P scan to support an optional correction step for EPI distortion. Note that newer GE scanners (ADNI4) include correction of EPI distortion in their reconstructions.
ADNI recommends a 2x2x2 voxel size.
This depends on the available scan time and/or whether the scanner is capable of 3x slice acceleration. At a minimum, ADNI recommends b = 0 and 1000 s/mm2 values be acquired, but for systems with multislice acceleration capability it also acquires b = 500 and 2000 s/mm2 shells, using custom diffusion vector sets. Acquiring multiple b shells enables analysis methods beyond DTI in order to more specifically separate microstructural changes from macrostructural ones (partial volume contamination of voxels by CSF due to atrophy).
MR vendors now apply their own pre-processing steps as part of the product, prompting ADNI to not perform its own pre-processing on images for ADNIs 3 and 4. ADNIs 1, and the initial parts of ADNI2/GO images were captured prior to vendor-initiated pre-processing, so the MRI Core provided intensity normalized and gradient un-warped T1 image volumes during these ADNIs. Later in ADNI2/GO, these corrections were applied on the scanner at the time of image reconstruction when these had been implemented in the vendor product.
ADNI has based their ASL sequence on recommendations from the “ISMRM White Paper”.
ADNI has found that 2D ASL imaging is by far inferior to 3D ASL imaging and has made the decision to no longer acquire 2D ASL data moving forward.