DicomRTTool 0.3.3

Tools for reading dicom files, RT structures, and dose files, as well as tools for converting numpy prediction masks back to an RT structure

This code provides functionality for turning dicom images and RT structures into nifti files as well as turning prediction masks back into RT structures

Installation guide

pip install DicomRTTool

Highly recommend to go through the jupyter notebook Data_Curation_and_Predictions_to_RT

Data_Curation_and_Predictions_to_RT has three main parts

Works on oblique images for masks and predictions*

1) Identify RT structures and names in multiple patients

2) Creating associations file and turning dicom into nifti/numpy files

3) Turning predictions into RT structures

If you find this code useful, please provide a reference to my github page for others www.github.com/brianmanderson , thank you!

Please consider using the .write_parallel if you have many patients

Ring update allows for multiple rings to be represented correctly

Various utilities created to help with the interpretation of dicom images/RT Structures

RT Structure and dicom conversion to numpy arrays

This code is designed to receive an input path to a folder which contains both dicom images and a single RT structure file

For example, assume a folder exists with dicom files and an RT structure located at 'C:\users\brianmanderson\Patient_1\CT1' with the roi 'Liver'

Assume there are 100 images, the generated data will be: Dicom_Image.ArrayDicom is the image numpy array in the format [# images, rows, cols]

You can then call it to return a mask based on contour names called DicomImage.get_mask(), this takes in a list of Contour Names

You can see the available contour names with

Example:

from DicomRTTool.ReaderWriter import DicomReaderWriter
Dicom_reader = DicomReaderWriter(description='Example')
path = 'C:\users\brianmanderson\Patients\'
Dicom_reader.walk_through_folders(path)  # This will walk through all of the folders, and print the indexes for each series instance UID present

all_rois = Dicom_reader.return_rois(print_rois=True)  # Return a list of all rois present, and print them



Contour_Names = ['Liver']
associations = {'Liver_BMA_Program4':'Liver','Liver':'Liver'}

Dicom_reader.set_contour_names(Contour_Names)
Dicom_reader.__set_associations__(associations)
Dicom_reader.get_images_and_mask()

image = Dicom_reader.ArrayDicom
mask = Dicom_reader.mask

pred = np.zeros([mask.shape[0],mask.shape[1],mask.shape[2],2]) # prediction needs to be [# images, rows, cols, # classes]
pred[:,200:300,200:300,1] = 1

output_path= os.path.join('.','Output')
Dicom_reader.prediction_array_to_RT(prediction_array=pred, output_dir=output_path, ROI_Names=['test'])

'''
Write the images and annotations as niftii files in parallel!
'''
Dicom_Reader.write_parallel(out_path=export_path, excel_file=os.path.join('.','MRN_Path_To_Iteration.xlsx'))