Preprocessing StrategyΒΆ
All data was preprocessed using CPAC. In order to streamline the preprocessing effort configuration files were made seperaely for FNIRT and for ANTs pipelines. The configuration files can be found here:
A development version of CPAC was used for preprocessing the release data (0.3.8_ahvo_scrubfix), however we recommend using the release version 0.3.9. These configuration files are compatible with both versions.
Pipeline Description
Apart from the variation in the 8 preprocessing pipelines, the strategies were kept the same. The following methods were used:
| Component | Method |
|---|---|
| Anatomical Registration | Performed in 1mm space for each of ANTs and FNIRT, with the transform being caluclated prior to skullstripping in the case of ANTs. |
| Tissue Segmentation | FAST was used for tissue segmentation, with the grey matter threshold set to 0.7, white matter threshold set to 0.96 and csf threshold also set to 0.96. |
| Functional Registration | Native functional to native anatomical registration was performed using boundary based registration. The mean functional was used as the template for the calculation. The functional was finally registed to 2mm MNI space. |
| Functional Masking | Brain Extraction Tool (BET) was used. |
| Nuisance Correction | The base nuisance correction strategy included linear detrending, quadratic detrending, motion correction, and the first 5 components of CompCor, an ICA based denoising technique. Added to this, for four of the pipelines, was global signal regression. |
| Temporal Filtering | Performed with a high pass cutoff of 0.01Hz and a low pass cutoff of 0.1 Hz. |
| Motion Correction | Fristons 24 Parameter Model was used for volume realignment. The scrubbing threshold, when used, was 0.2mm. |
| Spatial Smoothing | Kernel with full width, half maximum of 6mm. |
Derivatives
The following derivatives are also available:
- fALFF
- ReHo, with clusters sized to include 27 neighboring voxels.
- Dual Regression
- VMHC
- Network Centrality: Degree Centrality (weighted and binarized) with a sparsity threshold of 0.05 and Local Functional Connectivity Density (weighted and binarized) with a correlation threshold of 0.001.
