The EPAD Neuroimaging dataset

The EPAD imaging protocol is composed of some compulsory (core) sequences that are performed on all participants with the addition of other (advanced) modalities which the sites can optionally implement.

Core imaging Protocol

The core image acquisitions are mandatory and conducted in all sites. They provide structural and morphological information about the brain and serve to evaluate participants study eligibility through baseline radiological assessment and for quantitative analysis of brain structure and vascular lesions. Core sequences include:

• 3D-T1 weighted structural imaging(T1w)
• 3D Fluid Attenuated Inversion Recovery imaging (FLAIR)
• 2D-T2 weighted imaging (T2w)
• 2D-T2-star imaging (T2*)

Advanced imaging Protocol

The advanced Magnetic Resonance Imaging (MRI) protocol is performed only in a subset of sites with suitable equipment. It investigates functional and structural brain characteristics and includes one or more of the following types of acquisition:

• 3D-T2*
• 3D-susceptibility weighted imaging (SWI)
• Diffusion-weighted imaging (DWI or dMRI)
• Resting-state functional MRI (rs-fMRI)
• Arterial spin labelling (ASL)

Accessing Raw and Pre-processed Data

The original EPAD data from MRI scans in the form of DICOM data represents a high-volume data resource. To facilitate imaging research, the neuroimaging dataset is available in NifTI format and access will be provided via an xNAT currently hosted at VUmc (Netherlands).

All requests for EPAD LCS MRI data will be recorded by the EPAD LCS Research Access Process. Once a “data+imaging” request is approved, investigators will be given credentials to access the study portal on the xNAT system. Data in xNAT are stored following a standard Brain Imaging Data Structure (BIDS), and two types of NIfTIs can be accessed: raw and pre-processed.

Users interested in unprocessed data can refer to the “raw” folder on the xNAT system. This will include all NIfTIs that are direct results of the DICOM to NIfTI conversion and without any other manipulation applied. Each NIfTI is further accompanied with a JavaScript Object Notation (JSON), an open standard file format, storing information about image acquisition such as date and time, effective echo spacing, encoding direction, magnetic field strength and others. Given the high diversity of the EPAD dataset in terms of different scanner vendors and devices, JSON files are fundamental for extracting useful information for image pre-processing. Nonetheless, we strongly recommend users to aim for processed data, which will give the investigators the guaranty of working with an high-quality selected dataset.

Pre-processed data are stored in the “derived” folder, including all products which are not the direct results of the DICOM to NIfTI conversion. An overview of which pre-processed sequences can be found in xNAT is given in the table below.

Raw and pre-processed NIfTIs

Group	Sequence	Raw NIfTIs	Pre-processed NIfTIs
Core Sequences	3D-T1 weighted	✔	✔
	3D FLAIR	✔	✔
	2D-T2 weighted	✔
	2D-T2 star	✔
Advanced Sequences	3D-T2*	✔
	3D-SWI	✔
	DTI	✔	✔
	fMRI	✔	✔
	ASL	✔	✔

The Imaging Derivatives

The imaging derivatives are numerical variables which are extracted from different MRI sequences during pre and post processing. Such derivatives will be released in excel spreadsheets together with other EPAD data through the Aridhia workspace.

The EPAD imaging derivatives are thought for researchers with no or little experience in Image processing/analysis that want to carry out straight forward statistics on association between brain characteristics and other phenotypical data.

The table below shows the derivatives that are currently available (or in progress).

	Image Derivative	MRI Sequence	Description
Core Sequences	Leap Brain Volumes	T1	Volume of brain parcellations as computed with LEAP pipeline
	Freesurfer Parcellation	T1	Volume of brain parcellation as computed with Freesurfer
	White Matter Hyperintensities Volumes	FLAIR	Volume of White Matter Hyperintensities
Advanced Sequences	Bold Timeseries	fMRI	fMRI timeseries for different brain atlases (e.g. AAL, Harvard-Oxford)
	Functional Network Properties	fMRI	Graph properties (Degree, betweenness centrality, Eigenvector centrality, clustering Path length)
	Tract Based Spatial Statistics (TBSS)	DTI	Values of Fractional Anisotropy in the white matter skeleton

Besides the one illustrated in the table, the radiological read of structural scans provides the user with other information regarding brain pathology such as: number of microbleeds, atrophy scores (MTA, Koedam and GCA), and Fazekas score, global and per lobe.