Radiomics

Introduction

Radiomics involves reducing the image within the region of interest (anaotmical structure) to scalar features. This typically involves accumulating contributions from all the voxels within the region of interest; for example the mean of intensities from all the voxels. The radiomics features can be calculated for the original or the texture (or pre-processed) images.

Features

• First order (intensity)
• Peak and valley
• Shape
• Intensity Volume Histogram (IVH)
• Higher order (texture)

Signatures

Feature calculation used to generate various radiomics signatures is provided. For example, the following code computes the features used in Aerts et al's radiomics signature:

% Structure index
structNum = 1;

% Access the planC object
global planC

% Calculate the signature features
featureS = getNcomms5006Feature(structNum,planC)

Inter-software test suite

The radiomics test suite tests CERR's radiomics feature calculation with open-source software ITK and pyradiomics. Tests are provided for scalar features as well as pre-processing filters such as Wavelet and Laplacian Of Gaussian.

Examples

Wrapper function for feature calculation

global planC
scanNum = 1;
structNum = 1;
featS = calcGlobalRadiomicsFeatures(scanNum, structNum, paramS, planC);

calcGlobalRadiomicsFeatures is the wrapper function for calculating all types of global Radiomics features. It requires the follownig input parameters:

• scanNum: The index of the scan or the texture map (stored as pseudo scan).
• structNum: The index of the structure number (as in planC{indexS.structures}).
• paramS: The parameters used for feature calculation.
• planC: CERR's planC object. Can be accessed by global planC or by loading the CERR file in Matlab.

Obtain scan values within an ROI for the patient cohort: This is an example to obtain the scan values from each and every voxel within the region of interest for the cohort of patients. This can be useful to determine the optimal number of discretization levels required for feature calculation.

Batch calculate Global Radiomics features: This is an example to calculate global radiomics for a cohort of patients. It assumes that the anatomical structure is consistently named across the entire cohort.

Convert the output of calcGlobalRadiomicsFeatures.m to an analyzable matrix: The calcGlobalRadiomicsFeatures.m wrapper outputs the features as a Matlab structure array. This example shows how to convert this output to a matrix suitable for statistical analysis.

Citing CERR radiomics:

Apte AP, Iyer A, Crispin-Ortuzar M, Pandya R, V.van Dijk L, Spezi E, Thor M, Um H, Veeraraghavan H, Oh JH, Shukla-Dave A, Deasy JO (2018). Technical Note: Extension of CERR for computational radiomics: a comprehensive MATLAB platform for reproducible radiomics research, Med. Phys., 45 (8), pp. 3713–3720, August 2018. PMID: 29896896