fMRI-Preprocessing

Shell-based fMRI preprocessing and denoising pipelines used to compare the effect of different noise-removal strategies on the identification of resting-state networks (RSNs).

The repository implements multiple preprocessing and denoising variants based on motion regression, ICA-based cleanup (ICA-FIX), physiological signal regression, and their combinations.

Preprocessing outputs are used for group- and subject-level ICA analyses, followed by template matching with standard RSN's templates (Smith et al., 2009; Yeo et al, 2011).

---

Pipeline overview

The workflow is organized into three main stages: dataset configuration, preprocessing and denoising, and RSN extraction and evaluation.

1. Dataset and pipeline configuration

• Edit settings_dataset.sh to define dataset-specific parameters.
• Define dataset, task, run and denoise pipeline in main1.sh and main2.sh.

2. Preprocessing and denoising

• Run main1.sh to perform initial (minimal) preprocessing.
• Classify unknown ICA components returned by ICA-FIX and edit the corresponding .txt file in the mel.ica output directory.
• Edit and run main2.sh to select and apply the denoising pipeline and registration options.

Supported denoising steps include:

• Minimal preprocessing baseline: - B0 distortion correction, - Motion realignment, - Spatial smoothing, - High-pass temporal filtering
• Motion realignment parameter regression
• Motion outlier regression
• ICA-based denoising using ICA-FIX
• Regression of cerebrospinal fluid (CSF) and white matter (WM) signals
• Single- and multi-stage nuisance regression strategies

3. ICA and RSN analysis

• Perform ICA:
  • Group ICA: perform_group_ica.sh (preferred)
  • Single-session ICA: perform_singlesession_ica.sh
• Match ICA components to reference RSN templates:
  • Group ICA: perform_group_ica2.sh (define the reference RSN template)
  • Single-session ICA: compute_dice.sh (define the reference RSN template)
• Extract RSN features:
  • Group ICA: extract_rsns_group_features.sh
  • Single-session ICA: extract_rsns_features.sh

All scripts are intended to be opened and edited to define input variables prior to execution.

---

Software dependencies

All preprocessing, denoising, registration, and ICA steps in this repository are implemented using tools from (FMRIB Software Library) and executed via shell scripting.