Data for: An analytical workflow for seed-based correlation and independent component analysis in interventional resting-state fMRI studies.

Published: 24-07-2020| Version 1 | DOI: 10.17632/3dgtnm8j2t.1
Bhedita Seewoo,
Alexander Joos,
Kirk Feindel


Rs-fMRI data was acquired from 62 healthy adult (6–8 weeks old, 150–250 g) male Sprague Dawley rats under the effect of combined medetomidine-isoflurane anaesthesia (TE 11 ms and TR 1500 ms). Pre-processing of fMRI data included reorienting the brain into left-anterior-superior (LAS) axes (radiological view), skull-stripping, and upscaling the voxel sizes by a factor of 10. Single-session ICA was then carried out in FSL/MELODIC with the Gaussian kernel filter set to a full-width half maximum (FWHM) of 6.25 mm, a temporal high pass filter cut-off of 100 s and motion correction. FSL/FIX was manually trained by hand-labelling ICA’s decomposition of 60 datasets into signal or noise based on each component’s time-course, frequency, and spatial map. The trained weights training.RData file is provided here. The de-noised fMRI images were then coregistered to their respective T2-weighted images and normalised to a Sprague Dawley brain atlas. Multi-subject temporal concatenation group-ICA as implemented in FSL/MELODIC was carried out to identify functional network templates (provided here). These RSNs can be viewed by overlaying them on the brain atlas supplied. The use of an out-of-sample functional atlas is not recommended for rodent studies if study-specific variations are present. For example, the type and dose of anaesthetic used alter the relative localisation and strength of connectivity within specific networks and even the presence of some RSNs.