Skip to content

Latest commit

 

History

History
104 lines (74 loc) · 4.2 KB

statistical_analyses_with_spm8.md

File metadata and controls

104 lines (74 loc) · 4.2 KB
Raw

How to Process MRI Images:

  1. Beginner's Guide to Command-Line Interface
  2. Installation Instructions
  3. Preprocessing T1 images
  4. Hippocampus Segmentation

Reading Materials and Lecture Slides

Statistical Analyses with SPM / MATLAB

Smooth Images

  • Select all the jacobian images that need to be smoothed.
  • Either use the default or adjust the FWHM (blur intensity). With too little blur, you won't have enough voxels to form a significant cluster and with too much blur, you might washout significant clusters.

Factorial Design Specification

  • Select an output directory.
  • Select statistical design (re: two-sample t-test).
  • Select Group 1 scans (preferablly the control group).
  • Select Group 2 scans.
  • Select an explicit mask.

Model Estimation

  • Select SPM.mat file that contains the design specification.

Contrast Manager

  • Select SPM.mat file for contrasts
  • Select New: T-contrast

Contrast 1

  • Set name to TD > TBI
  • Set T contrast vector to 1 -1
  • Select New: T-contrast

Contrast 2

  • Set name to TBI > TD
  • Set T constrast vector to -1 1

Results Report

  • Select SPM.mat file that contains the design specification.
  • Select New: Contrast Query
  • Set results title to TD vs. TBI
  • Set Contrast(s) to inf
  • Set Threshold type to none
  • Set Threshold to 0.001

View Results in BrainNet Viewer

  • Select general surface file like BrainMesh_ICBM152.nv
  • Select volume file (re: SPM contrast file, spmT_0001.img or spmT_0002.img)
  • Select layout to full view if you want a jpeg image, select layout to single view if you want a movie.

Extract Binary ROI

  • Select Results
    • Select contrasts... and follow prompts
    • Select the cluster from which you want to make a binary ROI
    • Select save > current cluster

Identify Brain Structures

# In MATLAB
[oneline,cellarray]=cuixuFindStructure([4 -40 19]);

# Use either code below to output the structures
oneline{1}
cellarray

Extract Values

cd <PathToData>
for var in $(ls sjac*); do
    c3d -verbose $var <binaryROI>.img -label-statistics \
    >> <outputDir>/results.txt
done

Homework

Due on Monday, October 27, 2014

Write a statistic method and result section comparing controls versus severe TBI participants. Here's some example text you can use to draft your results section:

Data were analyzed using Statistical Parametric Mapping 8 software (SPM8) running on MATLAB 2013b. Jacobian maps were smoothed with an 8-mm isotropic full-width at half maximum (FWHM) Gaussian kernel in order to improve signal-to-noise ratio. Using these smooth Jacobian maps, we performed voxel-wise statistical analysis between groups using a two-sample T-test in order to identify changes in brain regions. SPM t-contrast maps were generated using a threshold of p < 0.001. Significance levels for the t statistic were controlled by the false discovery rate (FDR) for multiple comparisons and we report clusters that exceeded a threshold of 100 voxels and correct cluster level at p < 0.05. Anatomical localization of the cerebral areas was defined by an expericed observer, using the MNI atlas. Table x presents the regions of significant regional reduction in severe TBI participants compared to controls. Compared to controls, the severe TBI group showed significant regional reduction in the parietal lobe, [...] (see Figure x). Table y presents the regions of significant regional expansion in severe TBI participants compared to controls. Compared to controls, the severe TBI group showed significant regional enlargement of the lateral ventricles, [...] (see Figure y).