RadiologyNET Toolset - Public

API system and modules for different types of DICOM image manipulation and processing.

Installation

1. Clone this repository.
2. cd into the directory where you cloned this repository. Make sure you're in the same folder where setup.py is located.
3. Run pip install .

Organization

Currently, the package is divided into two main subpackages:

1. The tools subpackage is the core of the radiologynet-toolset package, it contains various utilities such as:

   • Raw data reading & parsing,
   • Data visualization, statistics and analysis,
   • DICOM image extraction and conversion,
   • Extraction of useful features from diagnoses:
     • Stripping words to their roots using a stemmer,
     • Building word corpus,
     • Implementation of Doc2Vec, TF-IDF...
   • DICOM tag analysis and preprocessing:
     • Encoding/decoding DICOM tags,
     • Selection of useful DICOM tags,
     • Imputing BodyPartExamined,
     • Using MissForest from missingpy to impute missing data,

2. The learn subpackage is dedicated to machine learning:

   • Dataset, dataloader and training of image feature extraction models:
     • U-Net,
     • R2U-Net,
     • AttU-Net,
     • Convolutional Autoencoder - CAE,
   • Building autoencoders for DICOM tags.

Throughout all of the provided code, there was special care given to documenting it. Hence, most of the given code has docstrings attached to it.