In bbuchsbaum/rMVPA: Multivoxel Pattern Analysis in R

MVPA Analysis on Regions of Interest

suppressPackageStartupMessages(library(neuroim2))
##suppressPackageStartupMessages(library(devtools))
library(rMVPA)

If we have one or more pre-defined regions of interests (ROIs) we can use rMVPA to run a clasisifation/regression model in each region, yielding peformance measures for each region.

Generate a volumetric dataset with 100 observations and two classes

To generate a dataset we use the gen_sample_dataset function. We are creating a 4-dimensional neuroimaging dataset, with 6-by-6-by-6 spatial dimensions and 80 observations in the 4th dimension. These 80 observations are divided into 4 blocks, each consisting of 20 trials. The generated y variable is a factor with 2 levels ('a' and 'b'). the gen_sample_dataset function creates a list with two elements: an mvpa_dataset object (dataset) and an mvpa_design object (design). The first contains information about the data itself and the second contains information about the experimental design.

dset <- gen_sample_dataset(D=c(6,6,6), nobs = 80, blocks=4, nlevels=2)
print(dset)

Now we generate an arbitrary set of ROIs by assigning each voxel in the input mask a value ranging from 1 to 3.

mask <- dset$dataset$mask
nvox <- sum(mask)

region_mask <- neuroim2::NeuroVol(sample(1:3,replace=TRUE,size=nvox), space(mask), indices=which(mask>0))
table(region_mask)

The region_mask volume now defines a set of three ROIs where each ROI is defined by an integer value. Next, we can run a classification/regression model within each ROI using the run_regional function:

mod <- load_model("sda")
tune_grid <- data.frame(lambda=.01, diagonal=FALSE)
cval <- blocked_cross_validation(dset$design$block_var)
mspec <- mvpa_model(mod, dataset=dset$dataset, design=dset$design, tune_grid=tune_grid,crossval=cval)
res <- run_regional(mspec,region_mask)

run_regional produces several output values, including performance_table, which includes cross-validated performance metrics for each ROI:

print(res$performance_table)

To get fine-grained information, the individual predictions for each ROI and trial are included in prediction_table:

print(res$prediction_table)

bbuchsbaum/rMVPA documentation built on Feb. 1, 2024, 8:45 p.m.