Alzheimer’s Disease Neuroimaging Initiative

Morphometry

Dugyu Tosun

Dugyu Tosun
Center for Imaging of Neurodegenerative Diseases, University of California, San Fransisco - FreeSurfer

Nick Fox

Nick Fox
Institute of Neurology, University College London

Nick Fox’s team focuses on brain, ventricle, and hippocampal boundary shift integral (BSI) and template-based regional measures. The Fox lab uses automated template-based methods for region delineation of brain, ventricles, and hippocampus with QC. Updated templates allow for more anatomic variability including more diversity. A volume estimate is generated for reach scan, longitudinal volume changes are measured with symmetric k-means normalized BSI giving a rate of atrophy or change shown to be robust to contrast differences and artifacts and to produce consistently amongst the lowest unbiased sample size estimate for clinical trials in ADNI and other data sets.

White Matter Hyperintensity Volume & Infarct identification

Charles DeCarli

Charles DeCarli
Director of Alzheimer’s Disease Research Center, University of California, Davis

Charles DeCarli utilizes automated template-based methods to remove non-brain tissue from high resolution T1 images. These images then undergo denoising, bias correction, alignment to a minimum deformation template and subsequent gray/white/CSF segmentation using Bayesian inference and maximal likelihood iterative convergence. FLAIR images are similarly denoised and segmented via a Bayesian inference approach. The presence, number, size, and location of cortical and subcortical infarctions are manually recorded.

ARIA-H

Petrice Coswell Jeff Gunter

Petrice Cogswell and Jeff Gunter
Aging and Dementia Imaging Research (ADIR) Lab, Mayo Clinic

Petrice Cogswell and Jeff Gunter lead the ARIA-H outputs for ADNI4. The x, y, z coordinates of each CMB and siderosis in subject space are entered into a database along with anatomic locations from the AAL atlas. An MEGRE sequence will be acquired on systems capable of efficiently saving phase and frequency images. SWI and traditional T2* GRE (from the 20 msec echo) are constructed from the MEGRE sequences.

Quantitative Susceptibility Mapping

Petrice Coswell Matthew Senjem

Petrice Cogswell and Matthew Senjem
Aging and Dementia Imaging Research (ADIR) Lab, Mayo Clinic

Petrice Cogswell and Matthew Senjem uses the STI suite to process the 3D-MEGRE data and generate QSM maps. Affine registration parameters are computed between the T1w images and mean of the magnitude GRE images across echo times and are used to align the TIV mask with the magnitude and phase GRE images. Laplacian-based phase unwrapping is applied, and masking performed to remove non-brain voxels from the T1 segmentation. Sparse linear equations and lease squares method are applied to compute the QSM from the unwrapped, masked phase data.

Dilated Perivascular Spaces

Petrice Coswell Jeff Gunter

Petrice Cogswell and Jeff Gunter
Aging and Dementia Imaging Research (ADIR) Lab, Mayo Clinic

Petrice Cogswell and Jeff Gunter employ an automated algorithm developed by Dr. Gunter for quantification. Dilated PVS have similar contrast to CSF, and parameters from simultaneous probabilistic segmentation of T1w, FLAIR, and T2w are combined with image-specific 3D spatial priors favoring tubelike structures to detect dilated PVS in WM and the basal ganglia. T1w and T2w contrasts provide localization while the FLAIR contrast differentiates dilated PVS terminating in WMH areas.

Diffusion Tensor Imaging

Paul Thompson

Paul Thompson
Laboratory of Neuroimaging, University of Southern California

Paul Thompson uses FSL software to correct for geometric distortions due to motion, eddy current, and susceptibility artifacts. The core pipeline has been updated from ADNI3 to improve dMRI model fit and derived measure sensitivity to biological effects including LPCA denoising, Gibbs de-ringing, and NODDI and MAP-MRI estimation. dMRI scans, denoised with Riemannian methods, are then registered to a geometrically-centered mean image. The parcellated Mori DTI81 atlas is overlaid to computer average values of all dMRI indices in ROIs. Voxel-by-voxel and ROI analysis is performed of the following DTI measures to distinguish from controls: fractional anisotropy (FA), geodesic anisotropy (GA), mean diffusivity (MD), and parallel and transverse diffusivity (diffusion tensor eigenvalues), and novel “beyond-tensor” metrics (NODDI, MAP-MRI, DKI) that are highly sensitive to both amyloid levels and clinical measures. This group also computes standard “network” measures most sensitive to AD-related change that predict decline (Connectomics).

Task Free Functional MRI

Jeff Gunter

Jeff Gunter
Aging and Dementia Imaging Research (ADIR) Lab, Mayo Clinic

Jeff Gunter heads the numeric outputs generated by TF-fMRI. Preprocessing steps include de-spiking using AFNI’s 3dDespike program, slice-timing correction, two-pass realignment to mean EPI, T1 co-registration to the mean EPI image, transforming a template space atlas Functional Connectivity Atlas into subject space, intensity bias corrections, high-pass filtering, re-sampling the data onto the cortical mesh, denoising of the data using ICA-fix, and spatial smoothing. Subject space spatial-temporal dual regression (STR) is performed on preprocessed data within a multi-variate framework incorporating four default mode (DMN) subsystems of interest (ventral DMN, posterior DMN, anterior-ventral DMN, and anterior-dorsal DMN) with z-score scaling of parameter estimates of functional connectivity within and between the DMN subsystems and is used to create a single summary metric of network failure.

Arterial Spin Labeling

Dugyu Tosun

Dugyu Tosun
Center for Imaging of Neurodegenerative Diseases, University of California, San Fransisco

Dugyu Tosun leads the ASL branch of the study. Perfusion is a measure of CVD and reflects perfusion deficits attributable to neurodegeneration, mirroring information provided by FDG PET. The UCSD team uses the model recommended by the ISMRM Perfusion Study Group for quantification of cerebral blood flow (CBF) from ASL images. Probabilistic tissue segmentation from the 3D T1 image will be used to correct for brain atrophy and gray/white matter partial volume effects. Summary CBF metrics for each subcortical and cortical FS parcellation are reported.

Multiple Medial Temporal Lobe Subregions

Paul Yushkevich

Paul Yushkevich
Biomedical Image Analysis, University of Pennsylvania

Paul Yushkevich aims to generate segmentations of multiple MTL subregions: subfields CA1, CA2, CA3, dentate gyrus, and subiculum of the hippocampus; entorhinal cortex; parahippocampal cortex; and perirhinal cortex (split into Brodmann areas 35 and 36). Subregions are further parcellated along the anterior-posterior axis, and the volume and average thickness of each subregion are reported. Segmentations use ASHS, a multi-atlas technique optimized for MTL subregion segmentation. This information is used to report subregion-specific measures of change in volume over time.

MRI Numerical Data Sets