Clinics to date¶
Temporal derivative and slice timing¶
February 2, 2015
Notebook on temporal derivatives
References:
Eklund, A., Andersson, M., Josephson, C., Johannesson, M., and Knutsson, H. (2012). Does parametric fMRI analysis with SPM yield valid results?—An empirical study of 1484 rest datasets. NeuroImage 61, 565–578.
Jason Steffener, Matthias Tabert, Aaron Reuben, Yaakov Stern, Investigating hemodynamic response variability at the group level using basis functions, NeuroImage, Volume 49, Issue 3, 1 February 2010, Pages 2113-2122, ISSN 1053-8119, http://dx.doi.org/10.1016/j.neuroimage.2009.11.014. (http://www.sciencedirect.com/science/article/pii/S1053811909011987) Keywords: fMRI; Hemodynamic response; Delay; Brain imaging; Basis sets; Group analysis
Thinking about power in neuroimaging analyses¶
February 9, 2015
Notebook on power in neuroimaging
(No clinic February 16, Presidents’ day).
Evaluation of the statistical model¶
February 23, 2015
Notebook on comparing models
Matthew mentioned the Worsley average (global) F-test for testing if a
particular F-test is significant on average across voxels – and therefore
whether to drop or keep the regressors in the F test. There is some very old
MATLAB code to compute this test here that may
well need adapting to newer versions of SPM.
Within-subject registration and fieldmaps¶
March 9, 2015
Exploring some DICOMs from the gre_field_mapping sequence - notebook on
investigating fieldmaps.
Useful links:
- http://nipy.org/nibabel/coordinate_systems.html
- http://www.mccauslandcenter.sc.edu/CRNL/tools/fieldmap
- https://bitbucket.org/matthewbrett/spm-versions/src/master/spm12/toolbox/FieldMap/FieldMap_principles.man
Jezzard P & Balaban RS. 1995. Correction for geometric distortion in echo planar images from Bo field variations. MRM 34:65-73.
Optimizing designs, orthogonalizing regressors¶
March 23, 2015
Notebook on optimizing designs
Notebook on correlated regressors
Multiple comparison correction, non-parametric methods¶
April 6, 2015
Notebook on FDR
Useful paper on difference in performance between different FWE methods:
Nichols, Thomas, and Satoru Hayasaka. “Controlling the familywise error rate in functional neuroimaging: a comparative review.” Statistical methods in medical research 12.5 (2003): 419-446. http://www-personal.umich.edu/~nichols/FWEfNI.pdf
On choosing a connectivity metric¶
April 13, 2015
Derek Nee presented on How I Learned to Stop
Worrying and Use DCM.
Spatial normalization¶
April 20, 2015
We talked about:
- the importance of the image affine for registration - see http://nipy.org/nibabel/coordinate_systems.html for a tutorial;
- diffeomorphic non-linear image registration for spatial normalization:
We also referred to the paper we’ll be discussing next week.
Assessing normalization methods¶
April 27, 2015
A journal club on:
Klein, Arno, et al. “Evaluation of 14 nonlinear deformation algorithms applied to human brain MRI registration.” Neuroimage 46.3 (2009): 786-802.
This paper attempts to do an objective comparison of different spatial normalization methods including SPM, FSL and ANTS
The SyN method referred to in Klein’s paper is symmetric diffeomorphic normalization – part of the ANTS software package.
Another less well-known package that did well in the tests is ART.
We had a fruitful and robust discussion about:
- the overall purpose of spatial normalization; is to increase overlap of functionally homologous areas? (as MB believes); does it vary according to the aim of the study (as JB believes)? Should we constrain ourselves to thinking only of the structural image?
- how much weight we can put on border drawn by specialists in neuroanatomy;
- what is the role of functional data in spatial normalization?
- how can we reasonably assess whether spatial normalization has been successful in a way that is independent of the measures the different methods are optimizing in doing the match?
Seminar on diffusion imaging analysis¶
May 4, 2015
by Ariel Rokem. See: Seminar in diffusion imaging.
Hands-on normalization with ANTS¶
May 11, 2015
Arielle Tambini showed us how to use the ANTS registration tools.
Her presentation is here.
Journal club on FMRI noise and FMRI networks¶
May 18, 2015
Daniel Sheltraw will lead a discussion on the following paper:
Molly G. Bright, Kevin Murphy, Is fMRI “noise” really noise? Resting state nuisance regressors remove variance with network structure, NeuroImage, Available online 7 April 2015, ISSN 1053-8119, http://dx.doi.org/10.1016/j.neuroimage.2015.03.070. (http://www.sciencedirect.com/science/article/pii/S1053811915002669)
Journal club on false positives in FMRI¶
June 8, 2015
A journal club on this paper:
Eklund, A., Andersson, M., Josephson, C., Johannesson, M., and Knutsson, H. (2012). Does parametric fMRI analysis with SPM yield valid results?—An empirical study of 1484 rest datasets. NeuroImage 61, 565–578.
http://dx.doi.org/10.1016/j.neuroimage.2012.03.093 http://www.sciencedirect.com/science/article/pii/S1053811912003825
The paper suggests that SPM (and maybe other packages) can have an extremely high false positive rates for real data, especially data with short TR.
We also briefly looked at the follow up conference paper discussed in this NeuroSkeptic blog post.
The general conclusion is that family-wise false positive error rates for standard first-level (within-subject) FMRI designs seem to be far too high in real (resting state) data. We started off discussing the very simple SPM autocorrelation model, but is was not clear that this was in fact at fault, and it appears from the follow-up paper that FSL has a worse false positive problem although it has a more sophisticated autocorrelation model.
Effect of movement and spatial preprocessing on functional SNR¶
June 15, 2015
A presentation by Anwar Nunez-Elizalde on some analyses of signal to noise with and without spatial processing such as motion correction.
Anwar also talked about the large reduction in subject motion from the custom “headcase” head restraint the Gallant lab are using.
(No meeting on June 22nd)
Journal club on detection of motion in FMRI¶
June 29, 2015
We looked at:
Erik B. Beall, Mark J. Lowe, SimPACE: Generating simulated motion corrupted BOLD data with synthetic-navigated acquisition for the development and evaluation of SLOMOCO: A new, highly effective slicewise motion correction, NeuroImage, Volume 101, 1 November 2014, Pages 21-34, ISSN 1053-8119, http://dx.doi.org/10.1016/j.neuroimage.2014.06.038. (http://www.sciencedirect.com/science/article/pii/S1053811914005151)
We concentrated on the question of how well standard volumetric means of identifying motion-corrupted fMRI data volumes perform.
Journal club on reproducibility¶
July 6, 2015
This was a discussion of what we mean by reproducibility in neuroimaging. W went through the arguments in Killeen et al. and discussed if the Killeen “p-rep” proposal is a good one -or not- for neuroimaging studies.
Killeen, Peter R. “An alternative to null-hypothesis significance tests.” Psychological science 16.5 (2005): 345-353.
Discussion of model validation¶
July 13, 2015
We discussed the papers below. The intention was to have some time to think of the problem more generally in terms of prediction and beyond activation models - for instance for the ICA model.
Loh, Ji Meng, Martin A. Lindquist, and Tor D. Wager. “Residual analysis for detecting mis-modeling in fMRI.” Statistica Sinica 18.4 (2008): 1421. http://www.stat.columbia.edu/~martin/Papers/SS_resid.pdf
Luo, Wen-Lin, and Thomas E. Nichols. “Diagnosis and exploration of massively univariate neuroimaging models.” NeuroImage 19.3 (2003): 1014-1032. http://www.fil.ion.ucl.ac.uk/spm/doc/papers/LuoNichols.pdf
In the discussion, JB also mentioned:
Jernigan, Terry L., et al. “More “mapping” in brain mapping: statistical comparison of effects.” Human brain mapping 19.2 (2003): 90-95.
Using components of CSF signal for denoising FMRI¶
July 20, 2015
We discussed regressing out white matter and CSF components for denoising. The question is: how do we choose the number of components?
We discussed that as a model selection question.
Journal club on denoising regression models in FMRI¶
July 27, 2015
We discussed:
Kay, Kendrick N., et al. “GLMdenoise: a fast, automated technique for denoising task-based fMRI data.” Frontiers in neuroscience 7 (2013). http://journal.frontiersin.org/article/10.3389/fnins.2013.00247/abstract
OpenFMRI data format¶
August 10, 2015
JB presented a proposal for the new OpenFMRI data format (called “Brain Imaging Data Structure”) developed by the OpenFMRI folks at Stanford and INCF neuroimaging group. It is meant to be a standard that will ease sharing of FMRI / other brain imaging data and be easy to implement.
Journal club on reliability in functional connectivity¶
August 31, 2015
Dan Lurie led a discussion of:
Mueller, Sophia, et al. “Reliability correction for functional connectivity: Theory and implementation.” Human Brain Mapping (2015). http://onlinelibrary.wiley.com/doi/10.1002/hbm.22947/full
(September 7, Labor day, no meeting)
Journal club in characterizing motion artefacts¶
September 14, 2015
We discussed:
Power, Jonathan D., et al. “Methods to detect, characterize, and remove motion artifact in resting state fMRI.” Neuroimage 84 (2014): 320-341. http://www.sciencedirect.com/science/article/pii/S1053811913009117
To start, Matthew ran some nipy diagnostics on a Berkeley FMRI dataset, to show that the top 20 PCA components appear to have signal related to movement, before and after standard SPM motion correction.
Then we looked at Power et al’s nice 2D plots of FMRI runs, showing large changes across many voxels that appear to be due to movement. Standard regression models using motion parameters and CSF / WM signal don’t do a good job of removing these, but regressing global signal and difference in global signal does appear to work well in reducing the effect.
Discussion of ROI voxel selection¶
September 21, 2015
Anna walked us through a problem selecting activated voxels, using MATLAB.
Among the things we discussed was whether to select voxels on their effect size (top half of t-statistic), or effect size relative to variance (t-statistic).
Removing noise with constrained regression¶
September 28, 2015
MB started by showing some systematic noise plots on example data, as we saw in the Power et al 2013 paper a few weeks back.
Then we looked at whether it’s possible to use constrained regression to look for, and regress out, scans where there exists large amounts of noise variance shared across voxels. There is the (edited) notebook, rendered nicely:
You can also download the notebook.
We were also looking at the scikit-learn LASSO description.
Dan L pointed us to this reference:
Satterthwaite, Theodore D., et al. “An improved framework for confound regression and filtering for control of motion artifact in the preprocessing of resting-state functional connectivity data.” Neuroimage 64 (2013): 240-256.
Discussion of brainspell activation database¶
October 5, 2015
We discussed the brainspell database and web application.
The code for brainspell is at: https://github.com/r03ert0/brainspell
The website is at: http://brainspell.org - but you may need to view it using Firefox to avoid startup errors.
Permutation inference for the GLM¶
November 9, 2015
A discussion of “Permutation inference for the general linear model.” by Winkler et al http://www.ncbi.nlm.nih.gov/pubmed/24530839
Idiot’s guide to tractography¶
November 13, 2015
How good is tractography?¶
December 4, 2015
Colin Hoy lead us in a discussion of:
Thomas, C., Ye, F. Q., Irfanoglu, M. O., Modi, P., Saleem, K. S., Leopold, D. A., & Pierpaoli, C. (2014). Anatomical accuracy of brain connections derived from diffusion MRI tractography is inherently limited. Proceedings of the National Academy of Sciences of the United States of America, 201405672. http://doi.org/10.1073/pnas.1405672111
The main question arising is - what is the best we can possibly do with standard tractography techniques, and is that good enough?
Parametric inference methods for group studies¶
March 15, 2016
http://arxiv.org/pdf/1511.01863.pdf
“CAN PARAMETRIC STATISTICAL METHODS BE TRUSTED FOR FMRI BASED GROUP STUDIES?”
It’s the follow-up to a study we looked at a while back, showing a big excess of false positives across many datasets for single-subject designs. Here the authors did the same kind of thing looking at the more common case of false positives in group studies, and found that the standard cluster-size methods can go way wrong. But - permutation tests work very well.
Reproducibility of psychological science¶
March 22, 2016
http://science.sciencemag.org/content/349/6251/aac4716.abstract
Estimating the reproducibility of psychological science BY OPEN SCIENCE COLLABORATION; SCIENCE AUG 2015
About 37 percent of somewhat randomly selected papers in psychology could be replicated.
Some questions:
- How good an estimate of reproducibility is this?
- Is 37 percent good enough?
- If 37 percent is the right answer in psychology, what is the answer likely to be in brain imaging?
Discussion of anatomical and functional connectivity¶
April 5, 2016
A discussion led by Katarina Slama functional connectivity: its relation to anatomical connectivity, but also “what is the partial in partial correlation and partial coherence?”
During discussion Melissa Newton pointed us to:
de Schotten, Michel Thiebaut, et al. “Monkey to human comparative anatomy of the frontal lobe association tracts.” Cortex 48.1 (2012): 82-96.
JB referenced:
Thomas, Cibu, et al. “Anatomical accuracy of brain connections derived from diffusion MRI tractography is inherently limited.” Proceedings of the National Academy of Sciences 111.46 (2014): 16574-16579.
Introduction to Fourier analysis¶
April 11, 2016
Working through the meaning of the Fourier analysis, using only sines and cosines : http://matthew-brett.github.io/teaching/fourier_no_ei.html
Constrained regression for removing outlier scans in FMRI¶
April 26, 2016
Matthew presented some preliminary results from using constrained regression to remove the effect of outlier scans in FMRI.
This is the follow-up to the presentation last September:
The idea was (and is) to put lots of noise regressors into the design matrix, and use constrained regression to drop almost all the noise regressors, leaving only the ones that really explain the data (such as delta regressors for outlier scans).
The notebook MB showed is at :
with some more explanatory notes (by Jonathan Taylor) in:
A format for describing and sharing FMRI data¶
May 11, 2016
JB presented a proposal for standardizing the description of the fmri data to ease sharing and standard processings:
Replication of brian-behavior correlations¶
June 7, 2016
JB led a discussion of some recent attempts to replicate brain-behavior correlations:
Boekel, W., Wagenmakers, E.-J., Belay, L., Verhagen, J., Brown, S., and Forstmann, B.U. (2013). A purely confirmatory replication study of structural brain-behavior correlations. Journal of Neuroscience 12, 4745–4765. http://newcl.org/publications/boekel-cortex.pdf
Kanai, R. (2016). Open questions in conducting confirmatory replication studies: Commentary on Boekel et al., 2015. Cortex 74, 343–347. http://linkinghub.elsevier.com/retrieve/pii/S0010945215000787
Muhlert, N., and Ridgway, G.R. (2016). Failed replications, contributing factors and careful interpretations: Commentary on Boekel et al., 2015. Cortex 74, 338–342. http://linkinghub.elsevier.com/retrieve/pii/S0010945215000775
Boekel, W., Forstmann, B.U., and Wagenmakers, E.-J. (2016). Challenges in replicating brain-behavior correlations: Rejoinder to Kanai (2015) and Muhlert and Ridgway (2015). Cortex 74, 348–352. http://linkinghub.elsevier.com/retrieve/pii/S001094521500221X