fMRI-Pre-processing

fMRI Pre-processing on SPM12 BIDS format - Suitable for Both Univariate and Multivariate Analysis

fMRI Pre-processing Pipeline for BIDS Format Data

Overview

This repository contains a comprehensive fMRI pre-processing pipeline implemented in SPM12 and MATLAB, optimized for both univariate and multivariate analyses. The pipeline follows a modified version of the protocol described in:

Di, X. & Biswal, B.B. (2023). A functional MRI pre-processing and quality control protocol based on statistical parametric mapping (SPM) and MATLAB. Frontiers in Neuroimaging, 1:1070151. https://doi.org/10.3389/fnimg.2022.1070151

Dataset

This pipeline is specifically configured for pre-processing the openly available dataset:

Lepping, R. J., et al. (2016). Neural Processing of Emotional Musical and Nonmusical Stimuli in Depression. PLoS ONE, 11(6), e0156859. https://doi.org/10.1371/journal.pone.0156859

The dataset is hosted on OpenNeuro: https://openneuro.org/datasets/ds000171

Key Features

• BIDS-compatible: Fully supports Brain Imaging Data Structure format
• MVPA-ready: No spatial smoothing applied to preserve fine-grained spatial patterns
• Comprehensive QC: Extensive quality control checks at each processing step
• Dual-task support: Handles both music and non-music task conditions
• Automated batch processing: Processes multiple subjects and runs efficiently

Pipeline Steps

Pre-processing (P)

1. P0: Slice Timing Correction (Added to original protocol)

   • Corrects for temporal differences in slice acquisition
   • Uses task-specific slice timing parameters

2. P1: Anatomical Segmentation

   • Segments T1w images into GM, WM, CSF
   • Generates deformation fields for normalization

3. P2: Realignment

   • Motion correction using rigid-body transformation
   • Generates motion parameters (rp_*.txt files)

4. P3a: Skull Stripping

   • Creates brain-only images for improved coregistration
   • Uses tissue segmentations from P1

5. P3b: Coregistration

   • Aligns functional images to skull-stripped anatomical
   • Uses normalized mutual information cost function

6. P4: Spatial Normalization

   • Warps images to MNI152 standard space
   • 3×3×3mm isotropic voxels
   • No spatial smoothing (preserves spatial patterns for MVPA)

Quality Control (Q)

• Q1: Initial data checks (parameters, orientation, artifacts)
• Q2: Segmentation quality assessment
• Q3: Head motion analysis (framewise displacement)
• Q4: Coregistration accuracy check
• Q5: Normalization quality verification

Usage

Running the Complete Pipeline

Execute scripts in numerical order:

% Slice timing correction
P0_slice_timing_correction

% Anatomical segmentation
P1_Segmentation

% Realignment
P2_realingment

% Quality checks
Q3_Head_Displacement

% Skull stripping and coregistration
P3_a_Skull_stripped
P3b_co_registration

% Normalization
P4_Normalization_MNI

% Final quality checks
Q4_Coregistration_check
Q5_Normalization_Check

First-Level GLM Analysis

For GLM analysis of preprocessed data:

GLM_First_LEVEL  % Reads events from TSV files

Output Structure

Depression/
├── sub-*/
│   ├── anat/
│   │   ├── c1*.nii      # Gray matter segmentation
│   │   ├── c2*.nii      # White matter segmentation
│   │   ├── c3*.nii      # CSF segmentation
│   │   ├── m*.nii       # Bias-corrected anatomical
│   │   ├── ss_m*.nii    # Skull-stripped anatomical
│   │   └── y_*.nii      # Deformation field
│   └── func/
│       ├── a*.nii       # Slice-time corrected
│       ├── rp_*.txt     # Motion parameters
│       ├── mean*.nii    # Mean functional image
│       └── wa*.nii      # Normalized functional images
├── QC_images/           # Quality control outputs
└── GLM_Results/         # First-level analyses

Key Modifications from Original Protocol

1. Added slice timing correction as the first preprocessing step
2. No spatial smoothing applied - preserves spatial resolution for multivariate analyses
3. Skull stripping implemented for improved functional-anatomical coregistration
4. Task-specific processing for music and non-music conditions
5. Enhanced quality control with automated exclusion criteria

Motion Exclusion Criteria

• Maximum framewise displacement threshold: 1.5mm or 1.5°
• Automated flagging of high-motion runs
• Subject-level exclusion if >50% of runs exceed threshold

Citation

If you use this pipeline, please cite:

1. The original protocol:

   Di, X. & Biswal, B.B. (2023). A functional MRI pre-processing and quality control 
   protocol based on statistical parametric mapping (SPM) and MATLAB. 
   Frontiers in Neuroimaging, 1:1070151.
   

Notes

• This pipeline is optimized for MVPA/searchlight analyses
• For univariate analyses requiring smoothing, add a smoothing step after normalization
• Quality control images are saved automatically for manual inspection
• Processing time: ~15-20 minutes per subject (varies by system)

Support

For questions or issues, please open an issue on GitHub or contact [erdemtaskiran3557@gmail.com].