Tabular MR data

Data users are strongly encouraged to consult the MRI data section on the ADNI site. It provides detailed imaging protocols and links to relevant documentation. This section covers tabular data related to MRI, including:

  • Procedural information and acquisition metadata
  • Numerical analysis results derived from images

Note that this document does not address the storage, formatting, or any other aspect of working with MR images themselves, which are available in DICOM format. We are concerned strictly with tables that relate to the imaging data.

MR image acquisition

This subheading in the IDA includes tables and documentation related to image acquisition. The IDA provides useful supporting materials (e.g., diffusion gradient tables for different manufacturers in ADNI3). Three tables are most relevant here: two used across all ADNI phases and one used historically.

More detailed acquisition information can be found in the DICOM headers of the images themselves, which should be considered the primary source for scan-specific acquisition information.

MRI acquisition over time

MR imaging has been a core component of the ADNI data set since the beginning of the study. The protocols for MR acquisition have changed over time, with these variations corresponding to the phases of the overall study (with ADNIGO and ADNI2 sharing the same protocol).

A table summarizing differences in protocols across different phases of the study is presented below, duplicated from the table available on this page:

ADNI1 ADNIGO/2 ADNI3 ADNI4

  • Focused on consistent longitudinal structural imaging on 1.5T scanners using back to back identical T1- and a dual echoT2-weighted sequences.

  • One-fourth of ADNI1 participants were also scanned using essentially the same protocol on 3T scanners.

  • Imaging was performed at 3T with T1-weighted imaging parameters similar to ADNI1.

  • In place of the dual echo T2-weighted image from ADNI1, 2D FLAIR and T2*-weighted imaging were added at all sites.

  • Both fully sampled and accelerated T1-weighted images were acquired in each imaging session.

  • Advanced imaging was included depending on scanner manufacturer:

    • GE: diffusion (dMRI) imaging

    • Philips: task-free resting state functional MRI (TF-fMRI)

    • Siemens: arterial spin labeling (ASL) blood flow imaging.

  • Imaging was captured exclusively on 3T scanners

  • dMRI, TF-fMRI, and ASL were included on all participants in ADNI3 protocol with a few site-wise exceptions

  • A two-tiered approach was created to accommodate the range of scanners for ADNI3: ‘ADNI3 Basic’ and ‘ADNI3 Advanced’ acquisitions were used for TF-fMRI and dMRI.

  • Exclusively done on 3T scanners with a limit to scan duration to one hour.

  • MRI protocol allows for equivalent imaging sequences across vendor-platforms to maintain uniformity across all images captured for the study

  • includes 3D T1, 3D FLAIR, 3D T2 GRE, T2, DTI, ASL, and TF-fMRI imaging sequences.
ADNI1 Scans*,** ADNIGO/2 Scans** ADNI3 Scans ADNI3 Scans

  1. 3 Plane Localizer

  2. Structural Sagittal T1

  3. Structural Sagittal T1

  4. REPEAT

  5. B1 Calibration – Head Coil (if applicable)

  6. B1 Calibration – Body Coil (if applicable)

  7. Axial PD/T2 Dual Echo

  1. 3 Plane Localizer

  2. Structural Sagittal T1

  3. Accelerated Structural Sagittal T1

  4. rsfMRI (Philips systems only)

  5. Axial T2-FLAIR

  6. Axial T2-Star

  7. Axial ASL Perfusion (Siemens systems only)

  8. Axial DTI Scan (GE systems only)

  1. 3 Plane Localizer

  2. Accelerated Structural Sagittal T1

  3. Sagittal 3D FLAIR

  4. Axial T2 STAR

  5. Axial 2D/3D PASL

  6. Axial DTI

  7. Axial rsfMRI

  8. High Resolution Hippocampus Scan

  1. 3 Plane Localizer

  2. Accelerated Structural Sagittal T1

    1. Compressed Sensed Accelerated Structural Sagittal T1 (where applicable)

  3. Sagittal 3D FLAIR

  4. Sagittal 3D T2

  5. Axial 2D/3D ME T2 GRE

  6. DTI

    1. Axial DTI PA (MB if applicable)

    2. Axial DTI AP (MB if applicable)

  7. ASL

    1. 3D Multi-PLD pCASL

    2. 3D Multi-PLD PASL

  8. Axial rsfMRI (MB if applicable)

  9. High Resolution Hippocampus Scan

*Note that phantom Localizer and MPRAGE scans were performed as part of each imaging study in ADNI1 for the purposes of performance monitoring and calibration. These scans are also available for download, but should not be confused with participant scans.

** Note that gradient distortion correction was inconsistent across scanner manufacturers, models and software versions. Offline correction was performed on T1-weighted images as needed with the net result being 3D distortion correction.  These corrections were validated using the phantom images.   Corrected versions are available in NIfTI format.   DICOM versions are directly from the scanner and care should be taken before using those images, particularly for longitudinal analyses.  In ADNI-3 and ADNI-4 offline distortion correction was no longer necessary. 

Imaging technology has evolved over the course of the study, and the protocol has kept pace. Because of the multi-site nature of ADNI, images are collected on many scanner models from multiple manufacturers. Information on scanner-specific protocols can be found here.

These protocol and hardware/software variations are particularly relevant for investigators working with longitudinal MRI data. Further guidance is available on this page, under the ‘scanner changes’ subheading.

1.5T and 3T scan information (MRIMETA, MRI3META)

These tables provide a summary of scan acquisition metadata derived from DICOM headers. Note that the DICOM headers themselves are the authoritative source for scan-specific acquisition metadata.

The MRIMETA table contains information on 1.5T acquisition, while the MRI3META table contains information on 3T acquisition (thus the ‘3’ in the name).

Usage notes

All ADNI4 scans were acquired at 3T, thus the 1.5T metadata table contains no entries for that phase.

The EXAMDATE field in these tables is the scan date, not the clinical visit date. Do not confuse this field with EXAMDATE in clinical data.

The MMTRNDATE field indicates the date on which the scan was uploaded to the LONI IDA, not the date on which the scan was acquired.

The HAS_QC_ERROR field in MRI3META is an administrative flag indicating a QC issue in the case report form (CRF), not in the image itself. It has no relevance to image QC.

The FLDSTRNGTH field is categorical: 1 = 1.5T, 2 = 3T. This information is duplicated in FIELD_STRENGTH.

Missing values may appear as -4 or NA. Treat both as missing; consult the data dictionary for legacy codes.

MR subject inclusion (MRINCLUSIO)

This table was in use from ADNI1 to ADNI2, and indicates whether specified MR findings were present for a subject and, when present, whether those findings triggered exclusion from the study. E.g. mass lesions, extensive vascular pathology, or other unexpected intracranial abnormalities identified on the screening scan.

The INCLUSION field indicates whether or not the finding was exclusionary, and other fields of the table indicate the type of findings indicated. The data dictionary provides more specifics on the contents.

In later phases, related information was included in more general inclusion/exclusion tables under the ENROLLMENT heading.

MR Image Analysis

The tables in this section contain numerical results of quantitative analyses conducted on raw or processed MR images. These analysis data sets are produced by a number of different labs, and each is accompanied by a methods document detailing the procedures used to derive the measurements and how they may have varied over time.

Some analyses were performed in earlier phases but not in ADNI4 (e.g., tensor-based morphometry). For ADNI4, the following measures are available:

  • Morphometric measurements:

    • Cross-Sectional Freesurfer - regional cortical and subcortical volumes/thickness measures

    • Automatic Segmentation of Hippocampal Subfields (ASHS) - segmented medial temporal subfield volumes

  • Boundary Shift Integral - longitudinal estimates of volume change between time points

  • Difusion Tensor Imaging (DTI) summary measures - summary DTI measures (e.g., fractional anisotropy, mean diffusivity)

  • White Matter Hyperintensity (WMH) Volumes - automated white matter hyperintensity segmentation volumes

  • Quantitative Susceptibility Mapping - regional measures of magnetic susceptibility

  • Network Failure Quotient - summary metric of large-scale network failure derived from resting state fMRI

Additional numerical summaries are available for previous phases. This area also includes visual read findings (e.g., infarcts).

General usage notes

Most of these tables share a common structure: one row per scan, with metadata and QC followed by global and/or regional summary measures. The exact regions for which summary measures are provided, and the atlases and segmentation schemes used to derive those brain regions, can vary between different measures. Details for each table should be available in the corresponding methods doc.

All of these tables provide a reference to the image(s) used to derive these summary measures. The unique identifiers for images are internally called UIDs, and each table contains a field referencing this identifier (e.g., IMAGEUID in the DTI summary table).

Tables generally record both the image acquisition date (e.g., EXAMDATE or SCANDATE) and the processing date (e.g., RUNDATE). Use the appropriate date for longitudinal modeling.

Some of these data sets include one or more QC-related fields to indicate whether the derived measures are suitable for downstream use. The specifics of these fields varies between tables. Users should consult the data dictionary entries and/or documentation for table-specific details.

Cross-sectional and longitudinal FreeSurfer (UCSFFSX7, UCSFFSL)

This data set contains morphometric measurements (volume, surface area, cortical thickness, etc.) for brain regions based on FreeSurfer segmentation.

The most recent cross-sectional results are available in the ‘UCSF - Cross-Sectional FreeSurfer (7.x)’ (UCSFFSX7) table. These results were calculated using FreeSurfer version 7.

This cross-sectional pipeline does not account for within-subject repeated measures. Each image is segmented independently.

Longitudinal FreeSurfer results, which do account for within-subject variability, are also available for participants who have completed all of their scheduled imaging sessions. Consequently, results for ADNI3/4 are not currently available.

The most recent outputs from the longitudinal pipeline cover scans from ADNI1/GO/2 participants, processed using FreeSurfer version 5.1. The results can be found in the ‘Longitudinal FreeSurfer (5.1) - Final Run’ (UCSFFSL51) table.

Documentation is available on the IDA and reproduced below:

Investigators who are interested in working with this data should read the documentation in depth.

Note that several tables containing data derived from older FreeSurfer versions are also available for download. Measurements are not necessarily comparable across versions; however these tables may be of interest to investigators researching methods for harmonization.

Usage notes

Before using the volumetric measurements in this table, users should review the various QC fields. Further details of the QC process and the interpretation of the fields can be found in the ‘QC Guide’ section of the methods doc above.

DTI regional summary measures (DTIROI_MEAN, DTIROI_ROBUSTMEAN)

This data set contains summary measures derived from diffusion tensor imaging, an MRI protocol intended to evaluate the microstructure of white matter tracts based on the anisotropy and structural orientation of water molecules.

These measurements are captured in two tables: DTIROI_MEAN and DTIROI_ROBUSTMEAN, each of which contain regional averages of four DTI metrics: fractional anisotropy (FA), mean diffusivity (MD), radial diffusivity (RD), and axial diffusivity (AxD). The two tables differ in the statistical methodology used to obtain these averages, with values in the DTIROI_ROBUSTMEAN table calculated using Huber’s M-estimator.

Documentation is available on the IDA and reproduced below:

Usage notes

The VOLUMES_AP and IMAGEUID_AP fields are populated only when an additional diffusion scan with opposite phase-encoding direction is available, in which case they provide a reference that scan.

The naming scheme for the columns containing summary measures is ‘measure_region_L/R’, where ‘metric’ is one of AD (axial diffusivity), FA (fractional anisotropy), MD (mean diffusivity), and RD (radial diffusivity). The regions denoted by each abbreviation are given in the data dictionary, and L/R denotes the laterality of the structure.

ASHS volume data (ADNI_PICSLASHS)

This table contains high-fidelity volumetric measurements of medial temporal regions, generated by ASHS software. This method generally provides more precise hippocampal and medial temporal subfield segmentation relative to FreeSurfer, and may be preferable for studies in which the primary outcome is specific to those regions.

Documentation is available on the IDA and reproduced below:

Usage notes

Usage notes for this table are provided in the above document, and reproduced here for greater visibility:

  1. Take note of the QA columns, which describe image quality on a scale from 0 (worst) to 4 (best). These represent overall usability of the T2-weigthed images and we don’t recommend using volumetric data for any image with a QA rating <= 1. Common factors contributing to low QA score are motion artifacts and slab orientation and positioning errors.
  2. For extrahippocampal cortical regions ERC, PHC, BA35 and BA36, use a normalized volume which can be obtained by dividing the raw volumes by the number of slices in which the ROI appears. This is because the segmentation of these regions along the MRI slice direction is partial. See Yushkevich et al. (PMID: 25181316) for additional details.
  3. CA2 and CA3 measurements are noisy due to the small size of these subfields. We recommend using CA1, or a combined CA (1+2+3) in your analysis if you can.

Boundary Shift Integral (FOXLABBSI)

This data set contains boundary shift integral (BSI) measurements for brain and ventricular boundaries, which capture longitudinal within-subject volume changes relative to a coregistered baseline scan from the same participant. Brain volume, ventricular volume and hippocampal volume for individual scans are also available.

Two variants of the metric are calculated: Classical BSI, in which intensity is normalized by dividing out the mean intensity in the intersection region between baseline and follow-up scans, and KN-BSI, which accounts for variance in intensity between CSF, white matter, and grey matter.

Further details on pre-processing and calculation can be found in the corresponding methods doc, which is available on the IDA and reproduced below:

Usage notes

Calculating BSI requires at least two images per subject. The older reference image is indicated by LONIUID_BASE. If LONIUID_BASE is missing, no reference scan is available and BSI values will be missing.

Images are quality checked both cross-sectionally and longitudinally, the results of which are recorded in the QC_PASS and REGRATING fields, repsectively. please see the data dictionary and methods doc for details on interpreting these fields.

White matter hyperintensity volumes (UCD_WMH)

This table contains volumetric measurements of white matter hyperintensities (WMH). WMH are associated with cerebrovascular pathology and serves as a biomarker of vascular pathology.

WMH volumes are produced from high-resolution 3D T1-weighted and FLAIR images using a robust Bayesian segmentation pipeline. For full details please see the relevant methods doc, which is available on the IDA and reproduced below: