README.md

In kbolab/moddicom: Package for handling DICOM RT files

Moddicom

Radiomics toolbox for R

"Dinapoli N, et al. Moddicom: a complete and easily accessible library for prognostic evaluations relying on image features. Conf Proc IEEE Eng Med Biol Soc. 2015 Aug;2015:771-4. doi: 10.1109/EMBC.2015.7318476."

Table of contents

• Installation (Windows)
• Installation (Ubuntu)
• Installation (Mac)
• Usage
• Use case 1: Extract features for a single patient DICOM series
• Use case 2: Extract features for multiple patients DICOM series
• Use case 3: Extract features for multiple patients DICOM series with LoG filter
• Contributing
• Credits
• License

Installation (Windows)

1) Download and install R v3: https://cran.r-project.org/bin/windows/base/ 2) Download and install Rstudio: https://www.rstudio.com/ 3) Download and install RTools: https://cran.r-project.org/bin/windows/Rtools/ 4) Download and unzip DCMTK libraries: https://dicom.offis.de/dcmtk.php.en (look for “DCMTK - executable binaries” and choose you OS architecture) 5) Add the .\bin folder of the unzipped DCMTK libraries to the PATH system environment variable 6) Install moddicom from github

library(devtools)

install_github("kbolab/moddicom")

Installation (Ubuntu)

1) Download and install R v3: https://cran.r-project.org/bin/windows/base/ 2) Download and install Rstudio: https://www.rstudio.com/ 3) Download and unzip DCMTK libraries: https://dicom.offis.de/dcmtk.php.en (look for “DCMTK - executable binaries” and choose you OS architecture) 6) Install moddicom from github

library(devtools)

install_github("kbolab/moddicom")

Installation (Mac)

1) Download and install R v3: https://cran.r-project.org/bin/windows/base/ 2) Download and install Rstudio: https://www.rstudio.com/ 3) Download and unzip DCMTK libraries: https://dicom.offis.de/dcmtk.php.en (look for “DCMTK - executable binaries” and choose you OS architecture) 4) Install moddicom from github

library(devtools)

install_github("kbolab/moddicom")

Usage

Requisites for a single patient folder: The class geoLet allows to open a folder of a single patient. The folder must not have subfolders and it must cointain only DICOM files with extension .dcm In the folder there must be only one series of images (CT or MRI) and one RTStruct; optionally there can be one RTDose and/or one RTPlan. The examples below can be run on the test images at https://github.com/kbolab/testImagesModdicom

Use case 1: Extract features for a single patient DICOM series

Launch Rstudio.

Load Moddicom library

library(moddicom)

Create a geoLet class instance.

obj.geolet <- geoLet()

Load the patient image series:

obj.geolet$openDICOMFolder(pathToOpen= "./test/Images/PAT1")

moddicom will load one by one the images and file with the structures, and will align the geometries.

Typing:

obj.geolet$getROIList()

you will see the ROI names for that patient.

TO extract a particular ROI, for instance “GTV-1”, you just type

roiVoxels <- obj.geolet$getROIVoxels(Structure = "GTV-1")

This will return a “list” data type. Since for our puroposes we’re just interested in the ROI VoxelCube, we can extract that as follows:

GTV_voxelcube <- roiVoxels$masked.images$voxelCube

Now is a 3D matrix with coordinates [x,y,z] To visualize a particular slice (the fifth in this example):

image(GTV_voxelcube[,,5], col = grey.colors(256))

Extract the features with the "single patient" method:

features <- computeFeatures.geoLet(obj.geoLet = obj.geolet, ROIName = roi, feature.family=c("stat","morph","glcm","rlm","szm"))

Result is a list of 232 features, whose first element is features$F_stat.mean and equals -46.88272 in this example.

Use case 2: Extract features for multiple patients DICOM series

library(moddicom)

path <- "./test/Images"

roi <- "GTV-1"

features <- f.extractor.sing.par(path = path, ROIName = roi, feature.family=c("stat","morph","glcm","rlm","szm"))

Result is a 2 by 233 data.frame, whose first two column are features[,1:2]:

patID | F_stat.mean ------------- | ------------- ./test/Images/PAT1 | -46.88272 ./test/Images/PAT2 | -257.72518

Use case 3: Extract features for multiple patients DICOM series with LoG filter

library(moddicom)

path <- "./test/Images"

roi <- "GTV-1"

features <- f.extractor.pluri.LoG.par(path = path, ROIName = roi, from.sigma = 0.6, to.sigma = 1.0, def.by = 0.2, feature.family=c("stat","morph","glcm","rlm","szm"))

Result is a length 3 list (one for each sigma value) of 2 by 233 data.frame. The first two column for the 3 dataframes are reported below:

features$`0.6`[,1:2]

patID | F_stat.mean ------------- | ------------- ./test/Images/PAT1 | -11.22115 ./test/Images/PAT2 | 56.40802

features$`0.8`[,1:2]

patID | F_stat.mean ------------- | ------------- ./test/Images/PAT1 | 12.7684 ./test/Images/PAT2 | 261.2649

features$`1`[,1:2]

patID | F_stat.mean ------------- | ------------- ./test/Images/PAT1 | 211.4133 ./test/Images/PAT2 | 1954.8270

Contributing

Credits

"Dinapoli N, et al. Moddicom: a complete and easily accessible library for prognostic evaluations relying on image features. Conf Proc IEEE Eng Med Biol Soc. 2015 Aug;2015:771-4. doi: 10.1109/EMBC.2015.7318476."

License

kbolab/moddicom documentation built on Nov. 29, 2020, 9:11 p.m.