xiaochengcike / radiomics-develop

A repository to develop code for IBSI compliant radiomics applications

Welcome to radiomics-develop!

This is a repository dedicated to the development of programming code for new radiomics applications (IBSI compliant!) via collaborative work with other scientific colleagues. I look forward to work with you! Please do not hesitate to contact me if you have any question: Martin Vallières, 1-514-839-0248, [email protected]

Recent developments

• IBSI calibration test -- Phase 2: Image processing with CT data (June 2018)
• IBSI calibration test -- Phase 1: Feature calculation with digital phantom (April 2018)
• Now using CSV files as inputs for roiNames (April 2018)
• Addition and subtraction of ROIs now allowed, with operations parsed in roiNames.csv files (April 2018)
• Compatibility with input imaging data of the following format: .nrrd and .img/.hdr (April 2018)
• Wavelet features (August 2017)
• Re-organization of radiomics structure (August 2017)
• Definition of definite/aboslute/filter imaging intensities (August 2017)
• Visualization tool to check ROIs on images with manualROIchoice=true (August 2017)

System requirements

• MAC or Linux operating systems. If you are using Windows, please install a Linux virtual box on your system.
• MATLAB 2017a and above.
• Python version 2.7 and above: "sudo apt-get install python" in linux.
• pydicom package: "sudo apt-get install python-dicom" in linux.

All software code have up to now been tested on:

• MATLAB 2017a using Ubuntu 16.04 (Linux).

Installation

1. Install git on your computer. In Linux, this goes simply by running sudo apt-get install git in the terminal.
2. Clone this online repository to your prefered location on your computer by running git clone https://github.com/mvallieres/radiomics-develop in the terminal.

Instructions

• COMPUTATION OF RADIOMIC FEATURES: Please follow the instructions in the INSTRUCTIONS.txt file at https://github.com/mvallieres/radiomics-develop/tree/master/WORKSPACE_RadiomicComputation
• IBSI calibration test -- Phase 1: Feature calculation with digital phantom: Please follow the instructions in the INSTRUCTIONS.txt file at https://github.com/mvallieres/radiomics-develop/tree/master/Tests/IBSI_FeatureCalculation
• IBSI calibration test -- Phase 2: Image processing with CT patient data: Please follow the instructions in the INSTRUCTIONS.txt file at https://github.com/mvallieres/radiomics-develop/tree/master/Tests/IBSI_ImageProcessing
• Other applications to come (multivariable modeling) ...

Documentation

• Full list of features and image processing steps are defined by the Imaging Biomarker Standardisation Initiative.
• Example applications of how to use texture features extracted using multiple parameters can be found in M Vallières et al 2015 Phys. Med. Biol. 60 5471 and in M Vallières et al 2017 Sci. Rep. 7:10117

Warning

• By default, GLCM, GLRLM and NGTDM matrices are using distance corrections as originally defined in M Vallières et al 2015 Phys. Med. Biol. 60 5471. Resulting features for these matrices will thus be slightly different from the benchmarked features defined by the Imaging Biomarker Standardisation Initiative.
• Moran's I index, Geary's C measure and Global intensity peak features are at the moment disabled. The reason is that these features take too long to compute the way they are written now. We need to find a way to vectorize their computation or to use valid approximations with faster computation methods.
• Comments and definitions of input and output arguments in the different pieces of code are not integrated yet. This is a work in progress.

Future work

1. Multivariable modeling.
2. Image post-processing (PVE corrections for PET, intensity nonuniformity corrections for MRI, etc.)

DISCLAIMER

"I'm not a programmer, I'm just a scientist doing stuff!"

STATEMENT

Copyright (C) 2017 Martin Vallières
All rights reserved.