The Open-Access European Prevention of Alzheimer's Dementia (EPAD) MRI dataset and processing workflow

Abstract

The European Prevention of Alzheimer Dementia (EPAD) is a multi-center study that aims to characterize the preclinical and prodromal stages of Alzheimer's Disease. The EPAD imaging dataset includes core (3D T1w, 3D FLAIR) and advanced (ASL, diffusion MRI, and resting-state fMRI) MRI sequences. Here, we give an overview of the semi-automatic multimodal and multisite pipeline that we developed to curate, preprocess, quality control (QC), and compute image-derived phenotypes (IDPs) from the EPAD MRI dataset. This pipeline harmonizes DICOM data structure across sites and performs standardized MRI preprocessing steps. A semi-automated MRI QC procedure was implemented to visualize and flag MRI images next to site-specific distributions of QC features - i.e. metrics that represent image quality. The value of each of these QC features was evaluated through comparison with visual assessment and step-wise parameter selection based on logistic regression. IDPs were computed from 5 different MRI modalities and their sanity and potential clinical relevance were ascertained by assessing their relationship with biological markers of aging and dementia. The EPAD v1500.0 data release encompassed core structural scans from 1356 participants 842 fMRI, 831 dMRI, and 858 ASL scans. From 1356 3D T1w images, we identified 17 images with poor quality and 61 with moderate quality. Five QC features - Signal to Noise Ratio (SNR), Contrast to Noise Ratio (CNR), Coefficient of Joint Variation (CJV), Foreground-Background energy Ratio (FBER), and Image Quality Rate (IQR) - were selected as the most informative on image quality by comparison with visual assessment. The multimodal IDPs showed greater impairment in associations with age and dementia biomarkers, demonstrating the potential of the dataset for future clinical analyses.

Keywords: EPAD; Image analysis pipeline; Magnetic resonance imaging; Multi-modal data integration; Quality control.

Fig. 1

Image processing workflow in the EPAD study. DICOM = Digital Imaging and Communications in Medicine; NIfTI = Neuroimaging Informatics Technology Initiative; QC = quality control; IDP = Image-derived phenotypes.

Fig. 2

Schematic diagram of preprocessing steps. Left: the core sequences preprocessing pipeline performed on 3D T1w and 3D FLAIR scans; Right: the three common steps of the advanced sequences preprocessing pipeline. Abbreviations: EPI = Echo-Planar Imaging.

Fig. 3

Overview of the quality control workflow. QC features are computed in the feature estimation module and cover 5 image features domains. Feature distributions can then be interactively inspected between-sites (2A) and within-sites (2B). Single-subject scans can be opened by clicking on the scatterplots (2C).

Fig. 4

Consort diagram representing number of scanned and successfully processed sequences. Abbreviations: EPAD = European Prevention of Alzheimer’s Dementia; T1w = T1 weighted; FLAIR = Fluid attenuated inversion recovery; fMRI = functional magnetic resonance imaging; dMRI = diffusion magnetic resonance imaging; ASL = Arterial spin labeling.

Fig. 5

3D T1w derived phenotypes. Example association between core sequences derived data and age. A) FreeSurfer surface reconstruction of one 3D T1w image; B) Association of eight cortical regional volumes with age; C) CAT12 tissue segmentation of one 3D T1 scan output: green = cerebrospinal fluid, blue = white matter, red = gray matter; D) Association of total gray matter volume, as computed with CAT12 segmentation, with age.

Fig. 6

FLAIR derived phenotypes. A) Example FLAIR scan from one EPAD participant with relatively high lesion volume; B) Result of the white matter hyperintensities (WMH) segmentation using BaMoS; C, D) Lobes and layer atlases, respectively, used for regional WMH volume computation, methodological details are given in (Sudre et al., 2018); E, F) Effect of age on WMH frequency (expressed in percentage of increase in frequency, i.e. the proportion of lesion in a given region, per additional year of age) for male and female respectively.

Fig. 7

Resting-state fMRI derived phenotypes. A) Six group resting-state networks spatial maps from a low dimensional (20 independent components) melodic ICA; B) Scatter plots showing the non-linear relationship of mean within-network functional connectivity with age, grouped by clinical dementia rating (CDR) score. R values are computed as the Pearson correlation coefficients between the quadratic age term (age²) and mean within network connectivity values.

Fig. 8

Diffusion MRI-derived phenotypes. A) The FA skeleton as computed in the TBSS pipeline (upper row) and the skeletonized white matter atlas used to extract local FA values (bottom row); B) Association of mean global FA values with age and amyloid status (as defined in (Ingala et al., 2021)). C) Association of 4 regional FA values with age and amyloid status. Abbreviations: FA = Fractional anisotropy; TBSS = Tract based spatial statistics; WM = White Matter; Sp. = Superior; CC = Corpus Callosum; r = Pearson correlation coefficient.

Fig. 9

Arterial spin labeling IDPs. A) Mean CBF in the gray matter across 237 participants. B) CBF in the GM relationship with age and APOE e4 carriership. Abbreviations:CBF = Cerebral blood flow; GM = Gray matter; APOE = Apolipoprotein E.