Figure 1.
Ghosting artefact due to animal movement in segmented EPI -acquisition.
Example of ghosting artefact in a 2-shot GE EPI image in four axial slices (9–12) from a volume acquired during a period of animal movement.
Figure 2.
Animal movement and trial concatenation.
Relationship of jaw and body movements with instabilities of the fMRI data before a) and after b) trial concatenation of one sample run of animal M1. The black thick line indicates the centre of mass (CoM) shift in millimetres relative to the first volume in the time series (see also [3]), representing apparent brain movement. The thin grey line indicates the recorded signal from the jaw movement sensor, and the thin black line the signal from the body movement sensor. Trial periods are shaded in light grey. Most movement occurs between trials and causes large artefacts visible in the CoM shifts. Trial concatenation removes most volumes containing such artefacts.
Figure 3.
Timecourse of a surface voxel affected by distortion.
Effect of the various processing steps on a single voxel time course. The x-axis depicts time in volumes, the y-axis shows voxel intensity values. The voxel time course: a) before any processing; b) after trials have been concatenated, showing low frequency drifts and a stepwise change at the boundary between scanning runs (vertical black line) c) when artefact periods are interpolated between trials in order to reproduce the original temporal arrangement of volumes; d) after filtering of interpolated data and removal of interpolated segments; there is now less variance in the time course due to filtering, though a step between runs remains.
Figure 4.
Adaptive brain extraction algorithm.
Demonstration of the effects of the algorithm implemented in fMRI Sandbox on an example slice (#10) of the mean image of a sample dataset M1: a) before applying any mask or filter; b) after applying a filter to suppress matter outside the brain; c) after brain extraction, now containing intensity changes in the brain due to the prior filtering; d) after a mask with the extraction data derived from c) has been applied to the original dataset.
Figure 5.
Modelled fixation and stimulation regressors for one single trial. The blue line depicts the modelled regressor for pure fixation (6 s), the green line the modelled regressor for visual stimulation (2 s).
Figure 6.
a) A mask from a standard analysis with artefact regressors. Due to the strong intensity changes in many voxels over time, the mask image is seriously degraded. b) Mask image after the mask threshold value has been adjusted to include the whole brain. The mask does now cover not only the brain but all other tissue as well. c) A mask image resulting from running a standard analysis after our advanced preprocessing. The mask image now homogenously covers the whole brain. Slices 7–11 are depicted.
Figure 7.
Activation maps for animal M1 after the consecutive data processing steps.
All runs have been modelled separately. Activation maps show the contrast for visual stimulation vs. simple fixation, overlaid on 5 slices (7–11) of the mean EPI-images. Colour bars show T-values; voxel threshold p = 0.001 uncorrected (T = 3.11); all cluster sizes shown.
Figure 8.
Activation maps for animal M2 after the consecutive data processing steps.
All runs have been modelled separately. Activation maps show the contrast for visual stimulation vs. simple fixation, overlaid on 5 slices (2–6) of the mean EPI-images. Colour bars show T-values; voxel threshold p = 0.05 FWE corrected (T = 4.96), all cluster sizes shown.
Figure 9.
Proportion of suprathreshold voxels for animals M1 & M2.
Number of voxels above a threshold of T = 3.11 are shown for methods 1–5 are depicted. a) Animal M1. b) Animal M2. Method 1: Standard modelling with SPM realignment; Method 2: Standard modelling with SPM realignment and artefact regressors; Method 3: Concatenated data with SPM realignment. Method 4: Concatenated data with 2-step realignment; Method 5: Concatenated data with 2-step realignment and high-pass filtering within FSB.