5622
Comment:
|
7405
|
Deletions are marked like this. | Additions are marked like this. |
Line 31: | Line 31: |
* Richard, John, Sylvain: Why is the notch filter not working well at the beginning of the signal? | |
Line 34: | Line 33: |
* Francois: Check that the bandpass filter is always used with enough data (Hilbert transform, SSP) | |
Line 47: | Line 45: |
[CODE] [ONLINE DOC] | [CODE] |
Line 65: | Line 63: |
* John: Dipole scanning: Wrong orientations | |
Line 69: | Line 66: |
* John, Richard, Sylvain: How to normalize unconstrained sources ? | * John, Richard, Sylvain: How to deal with '''unconstrained sources''' ? * For the Z-score normalization? * For the connectivity analysis? http://neuroimage.usc.edu/forums/showthread.php?2401 * For the statistics? * Projection on a dominant orientation? |
Line 85: | Line 86: |
* Enable option "Hide edge effects" for Hilbert (after finishing #10) | * Francois: Enable option "Hide edge effects" for Hilbert (after finishing #10) |
Line 89: | Line 90: |
* Hilbert: Link back to the filters tutorial (after finishing #10) * Describe what happens with unconstrained sources |
* Francois: Hilbert: Link back to the filters tutorial (after finishing #10) == Tutorial 25: Difference == [ONLINE DOC] * Validate order of the processes: Difference sources => Zscore => Low-pass filter<<BR>>Section: [[http://neuroimage.usc.edu/brainstorm/Tutorials/Difference#Difference_deviant-standard|Difference deviant-standard]] |
Line 95: | Line 100: |
* Francois: Implementation of Brainstorm-only permutation tests * Francois: Keep edge effects map (TFmask) for the stats on time-frequency maps |
|
Line 97: | Line 105: |
* Francois: Add a section about unconstrained sources (after finishing #27) |
|
Line 98: | Line 108: |
* Everything | * Update: http://neuroimage.usc.edu/brainstorm/Tutorials/WorkflowGuide * Update number of pages |
Line 101: | Line 112: |
* Everything | * Not evaluated yet * Update number of pages |
Line 106: | Line 118: |
* Francois: Check the number of pages for each tutorial | |
Line 111: | Line 122: |
== Phantom tutorial (for validation) == * John: Prepare Elekta recordings to distribute on the [[Tutorials/PhantomElekta|tutorial page]] * Francois: Prepare the Elekta phantom anatomy * Francois: Prepare the analysis script * Francois: Edit the tutorial page |
|
Line 118: | Line 125: |
* Francois: Update all the tutorials (100 pages) | * Francois: Update all the tutorials (100+ pages) * Francois: Remove useless images from all tutorials |
Line 120: | Line 128: |
* Francois: Split list in columns | |
Line 126: | Line 135: |
== Parametric experiments == * '''A = B''': Parametric t-test for constrained sources. 1. '''Sources''': Compute source maps for each trial (constrained, no normalization) 1. '''First-level statistic''': Compute a t-statistic for the source maps of all the trials A vs B. * Process2 "Test > Compute t-statistic": no absolute values, independant, equal variance. * With a high number of trials (n>30), t-values follow approximately a N(0,1) distribution. 1. '''Low-pass filter''' your evoked responses (optional). [NO! SHOULD BE DONE BEFORE, BUT WHEN ? => SPLIT ANALYSIS IN TWO: ERP OR FREQUENCY/RS] 1. '''Rectify '''the individual t-statistic (we're giving up the sign across subjects). 1. '''Project '''the individual t-statistic on a template (only when using the individual brains). 1. '''Smooth '''spatially the t-statistic maps. 1. '''Second-level statistic''': Compute a one-sampled chi-square test based on the t-statistics. * Process1: "Test > Parametric test against zero": One-sampled Chi-square test * This tests for '''|A-B|'''=0 using a Chi-square test: X = sum(|t<<HTML(<SUB>)>>i<<HTML(</SUB>)>>|^2) ~ Chi2(N<<HTML(<SUB>)>>subj<<HTML(</SUB>)>>) * Indicates when and where there is a significant effect (but not in which direction). 1. After identifying the significant effects, you may want to know which condition is stronger:<<BR>>Compute and plot power maps at the time points of interest: '''average(Ai^2^) - average(Bi^2^)''' |
Introduction tutorials: Editing process
http://neuroimage.usc.edu/brainstorm/Tutorials
Redactors:
Francois Tadel: Montreal Neurological Institute
Elizabeth Bock: Montreal Neurological Institute
Reviewers [current reviewing status]
Sylvain Baillet: Montreal Neurological Institute [overview 1-14, no proofreading]
Richard Leahy: University of Southern California [validated 1-19]
John Mosher: Cleveland Clinic [not started]
Dimitrios Pantazis: Massachusetts Institute of Technology [validated 1-15]
Expected timeline (2016)
March: Tutorials #1-#21 (interface and pre-processing)
April: Tutorials #22-#24 (source estimation and time-frequency)
May-June: Tutorials #25-#28 (statistics and workflows)
July-Sept: Cross-validation of the pipeline and the results with MNE, FieldTrip and SPM
October: Presentation during a satellite meeting at the Biomag 2016 conference
Nov-Dec: Connectivity tutorial
Tutorial 10: Power spectrum and frequency filters
[CODE]
- Richard, John, Sylvain: Finish the evaluation of the band-pass filters used in Brainstorm
- Francois: Mark with an extended event the time segments that we cannot trust (edge effect)
- Francois: Add warning if recordings are not long enough
[ONLINE DOC]
Richard, John, Sylvain: Section What filters to apply?
Richard, John, Sylvain: Section Filters specifications
Tutorial 12: Artifact detection
[ONLINE DOC]
Beth: Adding examples of the different detection processes. Section Other detection processes
Tutorial 13: Artifact cleaning with SSP
[CODE]
Francois: Check the length needed to filter the recordings (after finishing #10)
Section SSP Algorithm
Tutorial 15: Import epochs
[ONLINE DOC]
Richard, Sylvain, John, Francois: Define recommendations for epoch lengths (after finishing #10)
Section Epoch length
Tutorial 20: Head modelling
[ONLINE DOC]
- John, Richard, Sylvain: In-depth review of this sensitive tutorial
John, Richard, Sylvain: Section References
Tutorial 22: Source estimation
[CODE]
- John: Inverse code: sLORETA
- John: Inverse code: NAI
- John: Inverse code: Mixed head models
John, Richard, Sylvain: How to deal with unconstrained sources ?
- For the Z-score normalization?
For the connectivity analysis? http://neuroimage.usc.edu/forums/showthread.php?2401
- For the statistics?
- Projection on a dominant orientation?
- Francois: Remove the warning messages in the interface
Francois: Call FieldTrip headmodels and beamformers
[ONLINE DOC]
John, Richard, Sylvain: Section Source estimation options
John, Richard, Sylvain: Section Advanced options
John, Richard, Sylvain: Section Equations
John, Richard, Sylvain: Section References
- John, Richard, Sylvain: Why are dSPM values 2x lower than Z-score ?
- John, Richard, Sylvain: In-depth review of this sensitive tutorial
Tutorial 24: Time-frequency
[CODE]
- Francois: Enable option "Hide edge effects" for Hilbert (after finishing #10)
[ONLINE DOC]
- Francois: Hilbert: Link back to the filters tutorial (after finishing #10)
Tutorial 25: Difference
[ONLINE DOC]
Validate order of the processes: Difference sources => Zscore => Low-pass filter
Section: Difference deviant-standard
Tutorial 26: Statistics
[CODE]
- Francois: Implementation of Brainstorm-only permutation tests
- Francois: Keep edge effects map (TFmask) for the stats on time-frequency maps
[ONLINE DOC]
- Francois: Add a section about unconstrained sources (after finishing #27)
Tutorial 27: Workflows
Update: http://neuroimage.usc.edu/brainstorm/Tutorials/WorkflowGuide
- Update number of pages
Tutorial 28: Scripting
- Not evaluated yet
- Update number of pages
Final steps
- Francois: Check that the script tutorial_introduction.m produces the same output as tutorials
- Francois: Check all the links in all the pages
Francois: Reference on ResearchGate, Academia and Google Scholar
http://neuroimage.usc.edu/brainstorm/Tutorials/AllIntroduction
[Additional important stuff]
Other analysis scenarios
- Francois: Update all the tutorials (100+ pages)
- Francois: Remove useless images from all tutorials
- Francois: Add number of pages in the tutorials
- Francois: Split list in columns
Connectivity
- Not documented at all
- Richard, Sylvain: Define example dataset and precise results to obtain from them
- Richard: How to assess significance from connectivity matrices
- Francois: Preparation of a tutorial
Parametric experiments
A = B: Parametric t-test for constrained sources.
Sources: Compute source maps for each trial (constrained, no normalization)
First-level statistic: Compute a t-statistic for the source maps of all the trials A vs B.
Process2 "Test > Compute t-statistic": no absolute values, independant, equal variance.
With a high number of trials (n>30), t-values follow approximately a N(0,1) distribution.
Low-pass filter your evoked responses (optional). [NO! SHOULD BE DONE BEFORE, BUT WHEN ? => SPLIT ANALYSIS IN TWO: ERP OR FREQUENCY/RS]
Rectify the individual t-statistic (we're giving up the sign across subjects).
Project the individual t-statistic on a template (only when using the individual brains).
Smooth spatially the t-statistic maps.
Second-level statistic: Compute a one-sampled chi-square test based on the t-statistics.
Process1: "Test > Parametric test against zero": One-sampled Chi-square test
This tests for |A-B|=0 using a Chi-square test: X = sum(|ti|^2) ~ Chi2(Nsubj)
- Indicates when and where there is a significant effect (but not in which direction).
After identifying the significant effects, you may want to know which condition is stronger:
Compute and plot power maps at the time points of interest: average(Ai2) - average(Bi2)