Processing the diffusion-weighted magnetic resonance imaging of the PING dataset

Abstract

Diffusion-weighted magnetic resonance imaging (dMRI) allows for the in-vivo assessment of anatomical white matter in the brain, thus allowing the depiction of structural connectivity. Using structural processing techniques and related methods, a growing body of literature has illustrated that connectomics is a crucial aspect to assessing the brain in health and disease. The Pediatric Imaging Neurocognition and Genetics (PING) dataset was collected and released openly to contribute to the assessment of typical brain development in a pediatric sample. This current work details the processing of diffusion-weighted images from the PING dataset, including rigorous quality assessment and fine-tuning of parameters at every step, to increase the accessibility of these data for connectomic analysis. This processing provides state-of-the-art diffusion measures, both classical diffusion tensor imaging (DTI) and more advanced HARDI-based metrics, enabling the evaluation not only of structural white matter but also of integrated multimodal analyses, i.e. combining structural information from dMRI with functional or gray matter analyses.

Structural connectome fingerprinting and age prediction in pediatric development: assessing voxel- and surface-based white matter connectivity

“…The data that support the findings of this study are available in NIMH Data Archive (NDA) at 10.15154/1519178, study ID #932 ([dataset] Al-Sharif et al, 2020). These data were derived from the following resources also available in NDA as Pediatric Imaging, Neurocognition, and Genetics (PING) study ID #2607.…”

“…Anatomical MRI data were processed initially through one common dMRI processing pipeline, Tractoflow-pve , before subsequent processing to generate surface- and voxel-based connectomes. Processing steps are briefly described below; full details are described in (Al-Sharif et al, 2020), which also provides links to the data and processed derivatives.…”

“…For our purposes, we modified the original Tractoflow pipeline to include the whole-brain mask and partial volume estimation (PVE) maps for GM, WM, CSF and subcortical GM from CIVET 2.1 (Ad-Dab’bagh et al, 2006) along with the initial dMRI input. The pipeline is openly available through the NIMH Data Archive and GitHub as Tractoflow-pve (Al-Sharif et al, 2020).…”

“…We execute the test-retest and fingerprinting analyses using same-session test-retest data for 560 subjects from a pediatric imaging dataset and the age prediction in 813 subjects from the same study. These data are preprocessed with Tractoflow-pve (Al-Sharif et al, 2020; Theaud et al, 2020), a state-of-the-art diffusion processing pipeline from the Sherbrooke Connectivity Imaging Lab (SCIL), after which the connectomic mapping is completed using two different strategies: voxel- or surface-based. Due to the surface-based method’s more nuanced tractogram construction, we hypothesized that its connectivity representations should provide improved results over voxel-based measures in accurately depicting structural connectivity.…”

Network-based spreading of grey matter changes across different stages of psychosis

“…First, the DWI data for each individual were corrected using current best practise 40 for eddy-induced current distortions, susceptibility-induced distortions, intervolume head motion, outliers in the diffusion signal 41 , within-volume motion 42 , and B1 field inhomogeneities. Probabilistic tractography was conducted using a standardised and reproducible pipeline 43 and a state-of-the-art optimisation procedure was used to reduce false positive connections and improve biological interpretability of connection strength 44 . For a detailed overview of DWI processing and optimisation, see the Supplement .…”