How to compare a synthetic CT and a CT?

You’d like to compare a CT and a synthetic CT you’ve produced from an MRI. Unless each DICOM object contains a way of distinguishing them in its description, the first step is to classify your data in different directories as follows:

In the patient folder, create a sub-folder called “reference”, in which you will place the DICOM files of the CT, RTstruct and generated Rtdose. In the “synthetic” folder, you place the Rtdose generated from the synthetic CT and the synthetic CT files.

The REG matrix file is needed to link the synthetic CT and the CT reference.

The following script loads patient data into a list. It is then necessary to identify which items in the list belong to the “reference” folder and which belong to the “synthetic” folder. After loading the patient, we will create a list containing two elements (reference and synthetic), each containing DICOM objects (RTdose, RTstruct and scanner).

Feel free to run the script “step by step” (by placing your cursor on a line, then pressing CTRL + ENTER if you’re using RStudio).

library(espadon)
pat.dir <- "C:/data/my patient"

# patient loading
patient <- load.patient.from.dicom(pat.dir, data = TRUE)

# You can view the DICOM objects you have loaded:
display.obj.links(patient)

# List of reference and synthetic objects
pat <- lapply(c("reference", "synthetic"), function(dirname){
  modality.idx <- (which(names(patient) == "T.MAT") + 1):length(patient)
  obj.list <- lapply(modality.idx,function(idx){
    patient[[idx]] [sapply(patient[[idx]], function(l) tolower(basename(l$file.dirname))) == dirname]
  })
  names(obj.list) <- names(patient)[modality.idx]
  return( obj.list)
})
names(pat) <- c("reference", "synthetic")

# transfer matrix between the two geometric reference frames
tmat <- patient$T.MAT

# delete unnecessary data to free up memory space
rm(patient)
gc()

You can then inspect the RoI you can study by displaying part of the table given by pat[[“reference”]]$rtstruct[[1]]$roi.info

pat[["reference"]]$rtstruct[[1]]$roi.info[, c("name","roi.pseudo")]

                       name           roi.pseudo
1                centering             centering
2         EXTERNAL CONTOUR       externalcontour
3                  Bladder               bladder
4                    Rectum               rectum
5                    Sacrum               sacrum
6                      PTV                   ptv

We will choose here to calculate the metrics (Mean Error, Mean Absolute Error, Mean Squared Error, Root Mean Squared Error) of the RoI “Sacrum”, “Rectum”, “Bladder ” and “PTV ”.

roi.to.analyze.name <- c("Sacrum","Rectum", Bladder ", "PTV "),
roi.to.analyze.idx <- select.names(pat[["reference"]]$rtstruct[[1]]$roi.info$roi.pseudo,
                                   roi.name = roi.to.analyze.name)

metrics <- err.metrics.from.roi(obj = pat$synthetic$ct[[1]], 
                                 obj.ref = pat$reference$ct[[1]],
                                 struct = pat$synthetic$rtstruct[[1]],
                                 roi.name = roi.to.analyze,
                                 roi.sname = NULL, roi.idx = NULL,
                                 T.MAT = tmat)

Note that you could also have calculated the metrics using roi.sname or roi.idx, or by mixing roi.name, roi.sname and roi.idx.

The 3 uses of the err.metrics.from.roi function below are equivalent:

metrics <- err.metrics.from.roi(obj = pat$synthetic$ct[[1]], 
                                obj.ref = pat$reference$ct[[1]],
                                struct = pat$reference$rtstruct[[1]],
                                roi.name = NULL,
                                roi.sname = c("sacr","rectum",
                                              "bladd", "ptv"), 
                                roi.idx = NULL, .MAT = tmat)
# or

metrics <- err.metrics.from.roi(obj = pat$synthetic$ct[[1]], 
                                obj.ref = pat$reference$ct[[1]],
                                struct = pat$reference$rtstruct[[1]],
                                roi.name = NULL, roi.sname = NULL, 
                                roi.idx = c (3,4,5,6), T.MAT = tmat)
# or 

metrics <- err.metrics.from.roi(obj = pat$synthetic$ct[[1]], 
                                obj.ref = pat$reference$ct[[1]],
                                struct = pat$reference$rtstruct[[1]],
                                roi.name = NULL, roi.sname = NULL, 
                                roi.idx = roi.to.analyze.idx, T.MAT = tmat)

You can inspect the metrics variable

metrics 

      ROI         ME       MAE       MSE      RMSE
1 Bladder  12.427748  17.52306   539.189  23.22044
2 Rectum    7.496449  33.16230  6120.671  78.23472
3 Sacrum  -67.590704 152.87903 51226.703 226.33317
4 PTV      19.384144  45.51664 16568.974 128.72053

then save it in a *.csv file in the patient directory with the instruction :

write.csv2 (metrics, file.path (pat.dir, paste(basename(pats.dir), "metrics.csv")), 
            row.names = FALSE)