roi.Rmd

---
title: "Visulising ROI-level brain data in R"
output: html_notebook
---

We will now move onto visualising region level data. Often the result of our neuroimaging analyses are values assigned to specific brain regions. These values can be a test statistic (e.g. T-score), a effect size, a measure of cortical thickness or volume or anything at all. A great package for visualizing these values on brains is `ggseg` and `ggseg3D`. We will go through both below. Thanks to Dr Athanasia Mowinckel (https://github.com/Athanasiamo) for creating these packages and provided great documentation and tutorials.  

Lets first install the packages: 

```{r, include = FALSE}
install.packages("ggseg")
install.packages("ggseg3d")
install.packages("remotes")
remotes::install_github("LCBC-UiO/ggsegSchaefer")
```

And then load them up:

```{r}
library("ggseg")
library("ggseg3d")
library("ggplot2")
library("ggsegSchaefer")
```

Calling the `ggseg` function alone will display the base graphic (geom) used by ggseg (left/right, medial/lateral images):

```{r}
ggseg()
```

There are many brain atlases available in `ggseg`, but we will just use the Desikan-Killany cortical atlas ('dk') for now, and colour by regions .
```{r}
ggseg(atlas = "dk", mapping=aes(fill=region))
```


Lets get rid of the legend, and stack the four images and change the thickness of the boundary around each region and add a title. Try fooling around with the colour and size parameters:

```{r}
ggseg(mapping=aes(fill=region), 
      position="stacked",
      colour="black", #colour of the boundary
      size=.8, #thickness of the boundary 
      show.legend = F) +
  ggtitle("Desikan-Killany cortical atlas") #Add a title to the plot
```


Now lets try colour the brain regions by some value which relates to our own input data. All atlases included with `ggseg` contain data.tables with all the regions included in the atlas. Lets first grab the all the regions in the atlas for both hemispheres. 

```{r}
base_atlas <- as.data.frame(na.omit(cbind(dk$data$region, dk$data$hemi)))
colnames(base_atlas) <- c("region", "hemi")
head(base_atlas,5)
```

For now we are just going to generate random numbers as input, but here you can use your own values.

```{r}
set.seed(1993) #set a random seed (good practice for reproducibility)
brain_volume <- rnorm(dim(base_atlas)[1]) #generate a random numbers for each roi in the atlas
base_atlas <- cbind(base_atlas, brain_volume)
ggseg(base_atlas, 
      mapping=aes(fill=brain_volume), 
      position="stacked",
      colour="black",
      size=.8,
      show.legend = T) +
  ggtitle("Lets plot some values on the brain!")

```


Lets change the colour pallet used:

```{r}
ggseg(base_atlas,
      mapping=aes(fill=brain_volume),
      position="stacked",
      colour="darkblue",
      size=.8,
      show.legend = T) +
  ggtitle("Lets plot some values on the brain!") + scale_fill_gradientn(colours=c("dodgerblue4","light blue","firebrick","goldenrod"))
```

We can also select specific regions which are of interest to you:

```{r}
regions <- data.frame(cbind(region=c("superior frontal","insula"), # region name
                            hemi=c("right","right"), # hemisphere
                            value=c(2.4,5.5))) # value assigned

ggseg(regions,mapping=aes(fill=as.numeric(value)),
      hemi= c("right"), 
      colour="white",size=.3) +
  scale_fill_gradientn(colours=c("dodgerblue4","light blue","firebrick","goldenrod"))

```

Try completing the code below to select some other regions and assign values to them:

```{r}
#HINT: you can view the names of the ROIs in this atlas by calling base_atlas$region
other_regions <- (cbind(region=c(), # region name
                            hemi=c(), # hemisphere
                            value=c())) # value assigned
ggseg(other_regions,
      mapping=aes(fill=as.numeric(value)), 
      hemi= c("right", "left"), 
      colour="white",size=.3) 

```



`ggseg` also allows us to visualize subcortial structures. The base image (geom) is a set axial and sagittal slice, and the regions correspond the the `aseg` atlas from freesurfer.

```{r}
ggseg(atlas = "aseg")
```

Lets map some values onto subcortial structure. You can also remove regions we do not want by assigning them a `NA` value.

```{r}
regions = data.frame(cbind(region=c("hippocampus","pallidum",  "putamen"),
                           value=c(3,2,1)))

ggseg(atlas = "aseg",regions,mapping=aes(fill=as.numeric(value)), view = "axial") 

```


A sister package related to `ggseg`, called `ggseg3d` allows us to visualize our ROI level data in interactive 3D. We will briefly go though some of the basic functionality. The atlas we will use is the Schaefer 7-network resting-state atlas.

```{r}
ggseg3d(atlas = "schaefer7_3d")
```

Lets remove some of the axis labels and set the camera to a angle we wont.

```{r}
plot <- ggseg3d(atlas = "schaefer7_3d")  
plot <- remove_axes(plot)
pan_camera(plot, "left medial") 
```


We can also view subcortical structures in 3D. 

```{r}
ggseg3d(atlas = "aseg_3d")  
```

Here is an example of function I created with advanced use of ggseg3d which selects 14 commonly used subcortial regions (Left-Thalamus, Left-Caudate, Left-Putamen, Left-Pallidum Left-Hippocampus, Left-Amygdala, Left-Accumbens-area, Right-Thalamus-Proper, Right-Caudate, Right-Putamen, Right-Pallidum, Right-Hippocampus, Right-Amygdala, Right-Accumbens-area). It allows users to assign a user-specifed value to each of these regions, and places the camera at a nice angle. 

Here is an example use of this function:

```{r, eval = FALSE}
source("make_ggseg3d.R") #load in the function
set.seed(1993)
volume <- sample(1:20, 14) #randomly generated data for 14 structures
make_ggseg3d(volume, 
             colour.pal = c("white", "firebrick"))
```