Skip to contents

Vertex-wise analysis with threshold-free cluster enhancement (fixed effect)

TFCE_vertex_analysis.Rd

Fits a linear model with the cortical or hippocampal surface data as the predicted outcome, and returns t-stat and threshold-free cluster enhancement (TFCE) statistical maps for the selected contrast.

Usage

TFCE_vertex_analysis(
  model,
  contrast,
  formula,
  formula_dataset,
  surf_data,
  nperm = 100,
  tail = 2,
  nthread = 10,
  smooth_FWHM,
  VWR_check = TRUE
)

Arguments

model

An N X P data.frame object containing N rows for each subject and P columns for each predictor included in the model

contrast

A N x 1 numeric vector or object containing the values of the predictor of interest. Its length should equal the number of subjects in model (and can be a single column from model). The t-stat and TFCE maps will be estimated only for this predictor.

formula

An optional string or formula object describing the predictors to be fitted against the surface data, replacing the model, contrast, or random arguments. If this argument is used, the formula_dataset argument must also be provided.

  • The dependent variable is not needed, as it will always be the surface data values.

  • The first independent variable in the formula will always be interpreted as the contrast of interest for which to estimate cluster-thresholded t-stat maps.

  • Only one random regressor can be given and must be indicated as '(1|variable_name)'.

formula_dataset

An optional data.frame object containing the independent variables to be used with the formula (the IV names in the formula must match their column names in the dataset).

surf_data

A N x V matrix object containing the surface data (N row for each subject, V for each vertex), in fsaverage5 (20484 vertices), fsaverage6 (81924 vertices), fslr32k (64984 vertices) or hippocampal (14524 vertices) space. See also Hipvextract(), SURFvextract() or FSLRvextract output formats.

nperm

A numeric integer object specifying the number of permutations generated for the subsequent thresholding procedures (default = 100)

tail

A numeric integer object specifying whether to test a one-sided positive (1), one-sided negative (-1) or two-sided (2) hypothesis

nthread

A numeric integer object specifying the number of CPU threads to allocate

smooth_FWHM

A numeric vector object specifying the desired smoothing width in mm

VWR_check

A boolean object specifying whether to check and validate system requirements. Default is TRUE.

Value

A list object containing the t-stat and the TFCE statistical maps which can then be subsequently thresholded using TFCE_threshold()

Details

This TFCE method is adapted from the 'Nilearn' Python library.

See also

RFT_vertex_analysis, TFCE_vertex_analysis_mixed, TFCE_threshold

Examples

demodata = readRDS(system.file('demo_data/SPRENG_behdata_site1.rds',
package = 'VertexWiseR'))[1:5,]
CTv = readRDS(file = url(paste0("https://github.com",
"/CogBrainHealthLab/VertexWiseR/blob/main/inst/demo_data/",
"SPRENG_CTv_site1.rds?raw=TRUE")))[1:5,]

TFCEpos=TFCE_vertex_analysis(model=demodata[,c("sex","age")], 
contrast=demodata[,"age"], surf_data=CTv, tail=1, 
nperm=5, nthread = 2, VWR_check=FALSE)
#> The binary variable 'sex' will be recoded with F=0 and M=1 for the analysis
#> smooth_FWHM argument was not given. surf_data will not be smoothed here.
#> Estimating unpermuted TFCE image...
#> Completed in 5.6 secs
#> Estimating permuted TFCE images...
#> 
  |                                                                            
  |                                                                      |   0%
  |                                                                            
  |==============                                                        |  20%
  |                                                                            
  |============================                                          |  40%
  |                                                                            
  |==========================================                            |  60%
  |                                                                            
  |========================================================              |  80%
  |                                                                            
  |======================================================================| 100%
#> 
#> Completed in 0.1 minutes 

#To threshold the results, you may then run:
#results=TFCE_threshold(TFCEpos, p=0.05, atlas=1)
#results$cluster_level_results

#Formula alternative:
#formula= as.formula("~ age + sex")
#TFCEpos=TFCE_vertex_analysis(formula=formula, 
#formula_dataset=demodata, surf_data=CTv, tail=1, 
#nperm=5, nthread = 2, VWR_check=FALSE)