Spaarc Pipeline for Automated Analysis and Radiomics Computing

Node Icon

Extracts 165 standardized radiomic features from a region of interest in a medical image. It incorporates a comprehensive set of 8 convolutional filters, adhering to the guidelines established by the Image Biomarker Standardization Initiative - IBSI (Zwanenburg et al., 2020; Whybra et al., 2024).

Feature Families

The following Feature families and corresponding number of baseline features are implemented in SPAARC, (Whybra, 2021).

Type	Family Name	Baseline Feat No.
Shape-based	Morphological	23
First-order	Intensity-Based Statistics Intensity Histogram Intensity-Volume Histogram	18 23 7
Texture	Grey Level Co-occurrence Matrix (GLCM) Grey Level Run Length Matrix (GLRLM) Grey Level Size Zone Matrix (GLSZM) Grey Level Distance Zone Matrix (GLDZM) Neighbourhood Grey Tone Difference Matrix (NGTDM) Neighbourhood Grey Level Dependence Matrix (NGLDM)	25 16 16 16 5 16

Inputs

Image

Medical image for radiomic analysis

Type: Image, Required, Single

Mask

Binary mask highlighting region(s) of interest in the medical image for radiomic analysis.

Type: Mask, Optional, Single

Outputs

Output

A table which contains the estimated radiomic feature values.

Type: Table

Interpolated Image

Re-sampled image with new voxel spacing

Type: Image

Filtered Image

Response map obtained from image filtering proceedure

Type: Image

Morphological Mask

Region-of-Interest mask with the same grid as the image

Type: Mask

Intensity Mask

Re-segmented mask based on image intensities of the unfiltered image.

Type: Mask

IVH Discretised ROI

Discretized reigion-of-interest image utilised in estimating intensity volume histogram features.

Type: Image

IH and Texture Discretised ROI

Discretized reigion-of-interest image utilised in estimating intensity histogram and texture features.

Type: Image

Settings

Configure

Load Config File Filepath

Path to .json file with SPAARC configurations to be imported.

Feature Families

Morphological Boolean

Extracts morphological features if selected

Intensity Statistics Boolean

Extracts Intensity-Based Statistics features if selected

Intensity Histogram Boolean

Extracts Intensity Histogram features if selected

Intensity Volume Histogram Boolean

Extracts Intensity-Volume Histogram features if selected

GLCM Boolean

Extracts Grey Level Co-occurrence Matrix features if selected

GLRLM Boolean

Extracts Grey Level Run Length Matrix features if selected

GLSZM Boolean

Extracts Grey Level Size Zone Matrix features if selected

GLDZM Boolean

Extracts Grey Level Distance Zone Matrix features if selected

NGTDM Boolean

Extracts Neighbourhood Grey Level Dependence Matrix features if selected

NGLDM Boolean

Extracts Neighbourhood Grey Tone Difference Matrix features if selected

Interpolation

Method Selection

Specifies the interpolation algorithm to be applied.

Values: None, Linear, Spline

New Voxel Size [mm] Decimal Number Array

Specifies the new grid spacing for image interpolation in millimeters. If three voxel dimensions are specified, a full 3D interpolation is performed. If two voxel dimensions are specified, interpolation is performed slice by slice.

Round to Nearest Integer Boolean

Rounds the interpolated image to the nearest integer values if selected.

Convolutional Filters

Dimensionality Selection

Specifies the dimension of the convolutional filter.

Values: 2D, 3D

Filter Type Selection

Specifies the type of convolutional filter to be applied to the image.

Values: None, Mean, LoG, Laws, Gabor, Separable Wavelet, Non-Separable Wavelet, Riesz-Transform

Padding Method Selection

Specifies the method for imputing pixel values near the image boundaries for the convolutional filtering procedure.

Values: Constant, Nearest, Wrap, Reflect

Mean Filter Parameters

Support [px] Integer

Size of filter kernel defined in voxel units.

LoG Filter Parameters

Scale [mm] Decimal Number

Standard deviation of the Gaussian defined in millimeters.

Truncate Decimal Number

Number of standard deviations at which truncate the filter.

Laws Filter Parameters

Kernel i Selection

Laws 1D kernel to be applied to the i^th direction.

Values: Edges: 3, Edges: 5, Level: 3, Level: 5, Ripple: 3, Ripple: 5, Spots: 3, Spots: 5, Waves: 3, Waves: 5

Kernel j Selection

Laws 1D kernel to be applied to the j^th direction.

Values: Edges: 3, Edges: 5, Level: 3, Level: 5, Ripple: 3, Ripple: 5, Spots: 3, Spots: 5, Waves: 3, Waves: 5

Kernel k Selection

Laws 1D kernel to be applied to the k^th direction.

Values: Edges: 3, Edges: 5, Level: 3, Level: 5, Ripple: 3, Ripple: 5, Spots: 3, Spots: 5, Waves: 3, Waves: 5

Rotation Invariance Boolean

Make fIltered image equivariance to global rotations and invariance to local rotations if selected.

Pooling Selection

Method for voxelwise orientation pooling over the elements of the equivariant representation.

Values: Average, Max

Energy Map Boolean

Generates texture energy image if selected.

Distance [px] Integer

Chebyshev distance, utilized to create an energy image.

Gabor Filter Parameters

Wavelength [mm] Decimal Number

Inverse of the frequency of the oscillations, defined in millimeters.

Aspect Ratio Decimal Number

Ellipticity of the filter support.

Orientation [rad] Decimal Number

Orientation of the filter in radians.

Rotation Invariance Boolean

Make fIltered image equivariance to global rotations and invariance to local rotations if selected.

Pooling Selection

Method for voxelwise orientation pooling over the elements of the equivariant representation.

Values: Average, Max

Separable Wavelet Filter Parameters

Family Name Selection

Separable wavelet filter family name.

Values: Haar, Daubechies , Symlet, Coiflet, Biorthogonal, Reverse Biorthogonal, Discrete Meyer, Gaussian, Mexican Hat, Morlet, Complex Gaussian, Shannon, Frequency B-Spline, Complex Morlet

Wavelet Selection

Separable wavelet filter name.

Values: db1, db2, db3, db4, db5, db6, db7, db8, db9, db10, db11, db12, db13, db14, db15, db16, db17, db18, db19, db20, db21, db22, db23, db24, db25, db26, db27, db28, db29, db30, db31, db32, db33, db34, db35, db36, db37, db38, sym2, sym3, sym4, sym5, sym6, sym7, sym8, sym9, sym10, sym11, sym12, sym13, sym14, sym15, sym16, sym17, sym18, sym19, sym20, coif1, coif2, coif3, coif4, coif5, coif6, coif7, coif8, coif9, coif10, coif11, coif12, coif13, coif14, coif15, coif16, coif17, bior1.1, bior1.3, bior1.5, bior2.2, bior2.4, bior2.6, bior2.8, bior3.1, bior3.3, bior3.5, bior3.7, bior3.9, bior4.4, bior5.5, bior6.8, rbio1.1, rbio1.3, rbio1.5, rbio2.2, rbio2.4, rbio2.6, rbio2.8, rbio3.1, rbio3.3, rbio3.5, rbio3.7, rbio3.9, rbio4.4, rbio5.5, rbio6.8, gaus1, gaus2, gaus3, gaus4, gaus5, gaus6, gaus7, gaus8, cgau1, cgau2, cgau3, cgau4, cgau5, cgau6, cgau7, cgau8

Kernel i Selection

Separable wavelet 1D kernel to be applied to the i^th direction.

Values: High-pass, Low-pass

Kernel j Selection

Separable wavelet 1D kernel to be applied to the j^th direction.

Values: High-pass, Low-pass

Kernel k Selection

Separable wavelet 1D kernel to be applied to the k^th direction.

Values: High-pass, Low-pass

Rotation Invariance Boolean

Make fIltered image equivariance to global rotations and invariance to local rotations if selected.

Pooling Selection

Method for voxelwise orientation pooling over the elements of the equivariant representation.

Values: Average, Max

Decomposition Level Integer

Wavelet decomposition level of the image.

Non-Separable Wavelet Filter Parameters

Wavelet Selection

Non separable wavelet filter name.

Values: Simoncelli

Bmap Level Integer

Band-pass response map level.

Riesz-Transform Filter Parameters

Applied Filter Selection

Name of convolutional filter to be applied to the image prior to Riesz transformation.

Values: Riesz-Transformed LoG, Riesz-Transformed Simoncelli

Scale [mm] Decimal Number

Standard deviation of the Gaussian defined in millimeters.

Truncate Decimal Number

Number of standard deviations at which truncate the filter.

Bmap Level Integer

Band-pass response map level.

Order Integer Array

Order of the Riesz-based image derivatives.

Re-Segmentation

Range Re-Segmentation Boolean

Removes voxels outside a certain range from the intensity mask if selected.

Range Decimal Number Array

Specifies the lower and upper limits of the re-segmentation range. Voxel values not within these limits are exluded from the intensity mask.

Outlier Filtering Boolean

Removes voxels defined as outliers from the intensity mask if selected.

Sigma Decimal Number

Specifies the number of standard deviations from the mean of the intensities, used to define voxels as outliers.

Intensity Volume Histogram Discretisation

Binning Method Selection

Specifies the intensity binning method for calculating intensity volume histogram features.

Values: None, Fixed Bin Size, Fixed Bin Number

Bin Size Decimal Number

Specifies the fixed bin size for calculating intensity volume histogram features.

Number of Bins Integer

Specifies the fixed number of bins for the IVH discretisation procedure.

Intensity Histogram and Texture Discretisation

Binning Method Selection

Specifies the intensity binning method for calculating intensity histogram and texture features.

Values: Fixed Bin Size, Fixed Bin Number

Bin Size Decimal Number

Specifies the fixed bin size for calculating intensity histogram and texture features.

Number of Bins Integer

Specifies the fixed number of bins for calculating intensity histogram and texture features.

Texture Analysis

Analysis Type Selection

Specifies directional texture matrices analysis dimension.

Values: 2D, 3D, Both

GLCM Aggregation Selection

Specifies directional texture matrices aggregation method utilised GLCM features calculation.

Values: Merged, Averaged, Both

GLCM Distance Integer

Specifies distance from central pixel utilised for GLCM features.

GLRLM Aggregation Selection

Specifies directional texture matrices aggregation method utilised GLRLM features calculation.

Values: Merged, Averaged, Both

GLRLM Distance Integer

Specifies distance from central pixel utilised for GLRLM features.

NGTDM Distance Integer

Specifies distance from central pixel utilised for NGTDM features.

NGLDM Distance Integer

Specifies distance from central pixel utilised for NGLDM features.

NGLDM Alpha Decimal Number

Specifies a non-negative integer coarseness parameter for calculating NGLDM features.

Export

Folder Directory Path

Output folder.

Config File Name Text

SPAARC configuration file name.

Debug Boolean

Exports SPAARC data to .mat file if selected.

File Name Prefix Text

Prefix for Debug Data File Name.

Output Settings

Merge Features Tables Boolean

Merges output list into a single table if selected.

Interpolated Image Boolean

Returns the interpolated image as an output if selected.

Filtered Image Boolean

Returns the filtered image as an output if selected.

Morphological Mask Boolean

Returns the morphological mask as an output if selected.

Intensity Mask Boolean

Returns the intensity mask as an output if selected.

IVH Discretised ROI Boolean

Returns the discretised ROI, ustilised in calculating intensity volume histogram features, as an output if selected.

IH and Texture Discretised ROI Boolean

Returns the discretised ROI, ustilised in calculating intensity histogram and texture features, as an output if selected.