FMRI#

Studying resources and introduction to fMRI

Important resource#

Following book and videos will lead and cover the entire studying of fMRI:

  • Textbook: Handbook of Functional MRI Data Analysis

  • Videos: Principles of fMRI

Meanwhile, AFNI, as one of the leading softwares for the analysis and display of multiple MRI modalities will be widely used in the lab. Click here to learn more and follow the instruction to install it on your local computer.

Practice example#

This plan will use a recent study of our lab as practices in fMRI data preprocessing, analysis and visualization. The study, Distributed and Multifaceted Effects of Threat and Safety, also named as MAX, aims to examine how anxious apprehension is processed in the human brain. Details of data and models can be found at our github page.

All data require access permission to Bwift@UMD.eud Raw data:

/data/bswift-1/Pessoa_Lab/MAX/dataset/raw/MAX???

Stimulus timing files

/data/bswift-1/Pessoa_Lab/MAX/stim_times_neutral/

Study plan#

1. Introduction to fMRI#

  • Charpter 1-2 in Textbook

  • Module 1-12 of Part 1 in Videos

Key points

  • Physical and chemical principle of MRI scan

  • What are structural and functional MRI images? What’s the difference?

  • BOLD signal and its features

  • Artifacts and noise of fMRI

  • Different types of experiment design

Addition to above readings and videos, here are some complementary contents which would be useful in this charpter:

See also

2. Preprocessing of fMRI data#

  • Charpter 3-4 in Textbook

  • Module 13-14 of Part 1 in Videos

Key points

  • Why we can’t use fMRI images directly gained from scanner?

  • What’re the general steps in preprocessing? What’s the meaning of each step?

The scripts for preprocessing:

data/bswift-1/Pessoa_Lab/MAX/scripts/all_preproc.sh

Caution

The real fMRI preprocessing will be quite time consuming for just one subject

The proprocessed data(for neutral runs):

  • Smooth:

/data/bswift-1/Pessoa_Lab/MAX/dataset/preproc/MAX???/func_neutral/MAX???_EP_Main_TR_MNI_2mm_SI_denoised.nii.gz

  • Unsmooth:

/data/bswift-1/Pessoa_Lab/MAX/dataset/preproc/MAX???/func_neutral/MAX???_EP_Main_TR_MNI_2mm_I_denoised.nii.gz

Complementary contents:

See also

3. First level analysis#

Before dive into this charpter, one should at least have some systemic awareness of probability and statistics. Thus, complete courses are recommended. However, here are some good textbooks and YouTube channels for quick reviewing and reinforcement:

See also

There are a lot of excellent resources other than above list. Please be free to search, ask and learn.

For fMRI related analysis:

  • Textbook: Charpter 5

  • Module 15-22 of Part 1 in Videos

Key points

  • The details of general linear regression and univariate approach for voxels/region of interest in a fMRI brain map

  • The choice of independent variables in the model

  • What is hemodynamic response function(HRF)? Why deconvolution?

  • The pros and cons of assumed and unassumed shape HRF

  • Multicollinearity

The scripts for first-level analysis:

  • voxel-wise

data/bswift-1/Pessoa_Lab/MAX/scripts/Murty_Final/voxelwise_analysis/condition_level/MAX_fMRI_Analysis_neutral_deconv_reducedRuns.sh

  • ROI-wise

/data/bswift-1/Pessoa_Lab/MAX/scripts/Murty_Final/ROI_analysis/condition_level/FNSandFNT/MAX_fMRI_Analysis_neutral_deconv_reducedRuns.sh

Reults of first-level analysis:

  • voxel-wise

/data/bswift-1/Pessoa_Lab/MAX/dataset/first_level/voxelwise/neutral_runs_conditionLevel_FNSandFNT/MAX???/

  • ROI-wise

/data/bswift-1/Pessoa_Lab/MAX/dataset/first_level/ROI/neutral_runs_conditionLevel_FNSandFNT/MAX_ROIs_final_gm_85/MAX???/

ROI-wise responses visualization

https://github.com/LCE-UMD/mood-anxiety/blob/main/01-final_gm_85/01-data_preprocessing.ipynb

4. Group level analysis#

  • Charpter 6-7, 10 in Textbook

  • Module 23-29 of Part 1 in Videos

Key points

  • Why group level analysis?

  • Generalized linear model, including parametric ones like t-test, ANOVA, ANCOVA, mixed effect model and non-parametric ones like permutation test, rank test etc.

  • P-value and types of errors

  • Correction for multipule comparsion

Useful AFNI program

5. What’ more#

Apart from above standard approach for fMRI analysis, more interesting contents such as brain connectivity, selection bias, graph theory etc. are just around the corner. The remaining part of Textbook and Part 2 of Videos have more details for them. Moreover, in addition to analyze fMRI images from frequentist point of view, Bayesian approach is gradually becoming popular. Here is a good notebook about practical Bayesian model in fMRI

Have fun in learning!

License#

LCE@UMD