Introduction to fMRI Analysis with SPM

Types of Data Collected

1. T1 Structural Data

   • High-resolution structural images.
   • ~1mm resolution allowing for detailed anatomical landmarks.

2. Functional Data

   • Lower resolution compared to structural data.
   • Acquired in real-time during a task.

3. Field Map Data

   • Records inhomogeneities in the magnetic field (e.g., air pockets from sinuses).

Pre-Processing Steps

1. Obtain Scanner Sequence Information

• From DICOM headers or printed PDFs from the scanner.
• Necessary for entering parameters into SPM.

2. Unwarp and Realign Data

• Uses field map to correct magnetic field inhomogeneities.
• Restoration of data positioning but not lost data.

3. Slice Timing Correction

• Functional images acquired in slices over time.
• Correcting for different time points of acquisition within a TR.

4. Co-registration

• Aligns functional and structural data into the same space.
• Ensures anatomical landmarks are aligned.

5. Segmentation

• Compares structural image to a standard brain template using tissue probability maps (gray matter, white matter, CSF).
• Generates deformation fields for normalization.

6. Normalization

• Aligns functional data to standard space using deformation fields.

7. Smoothing

• Averages activations over multiple voxels.
• Reduces noise in the data.

Accessing and Using Jupyter with SPM

Setting up PuTTY for Jupyter Access

• Host Name: “jupyter.mclean.harvard.edu”
• Enable X11 Forwarding under Connection > X11.

Launching MATLAB

• Set path to access files.
• Extract scanner parameters from DICOM headers using dicominfo command.

Using SPM for Pre-Processing

Launching SPM

• Download and add to MATLAB path.
• Use Batch Processor for sequential pre-processing steps.
• Run modules and save batch for future use.

Step-by-Step Pre-Processing in SPM

1. Field Map Preparation

   • SPM > Tools > Field Map > Calculate VDM
   • Specify phase image, magnitude image, and other parameters.

2. Realign and Unwarp

   • SPM > Spatial > Realign and Unwarp
   • Select functional images; use calculated VDM from the previous step.

3. Slice Timing Correction

   • SPM > Temporal > Slice Timing
   • Input TR, slice order, and reference slice.

4. Co-registration

   • Align mean functional image with structural image.

5. Segmentation

   • Apply tissue probability maps and save deformation fields.

6. Normalization

   • Use deformation fields to normalize the data.

7. Smoothing

   • Dependent on normalized images.

Checking Registration

• Ensure functional data aligns with the standard brain template.
• Validate the alignment of anatomical landmarks and edges.

Summary

• The pre-processing steps include obtaining scanner parameters, unwarping, realignment, slice timing correction, co-registration, segmentation, normalization, and smoothing.
• Using SPM tools to automate and streamline the pre-processing pipeline.
• Verifying the quality of registration by comparing functional data to standard brain templates.