Radiomics – Psychosensing Lab

Radiomics and radiomic features

Radiomics is a method that aims to extract a large number of features from medical images, using data-characterisation algorithms.

The set of features can be divided into a number of families, such as intensity-based statistical, intensity histogram-based, intensity-volume histogram-based, morphological features, local intensity, texture matrix-based features, etc. To calculate features, it is assumed that an image segmentation mask exists, which identifies the voxels located within a region of interest (ROI).

The image biomarker standardisation initiative (IBSI, [Zwanenburg 2020]) is an independent international collaboration which works towards standardising the extraction of image biomarkers (image features) from acquired imaging for the purpose of high-throughput quantitative image analysis (radiomics).

Example

Segmentation process (data from ACRIN-FLT-Breast. The Cancer Imaging Archive, [Kinahan, 2017]), using Slicer 3D software [Fedorov, 2012]:

Image type	Feature Class	Feature Name	Segmentation_segment_Segment_1	Segmentation_segment_Segment_2
diagnostics	Versions	PyRadiomics	v3.0.1.post4+gad5b2de	v3.0.1.post4+gad5b2de
diagnostics	Versions	Numpy	1.19.2	1.19.2
diagnostics	Versions	SimpleITK	2.0.2	2.0.2
diagnostics	Versions	PyWavelet	1.1.1	1.1.1
diagnostics	Versions	Python	3.6.7	3.6.7
diagnostics	Configuration	Settings	{’minimumROIDimensions’: 2, 'minimumROISize’: None, 'normalize’: False, 'normalizeScale’: 1, 'removeOutliers’: None, 'resampledPixelSpacing’: None, 'interpolator’: 'sitkBSpline’, 'preCrop’: False, 'padDistance’: 5, 'distances’: [1], 'force2D’: False, 'force2Ddimension’: 0, 'resegmentRange’: None, 'label’: 1, 'additionalInfo’: True, 'binWidth’: 25.0, 'symmetricalGLCM’: True, 'correctMask’: True}	{’minimumROIDimensions’: 2, 'minimumROISize’: None, 'normalize’: False, 'normalizeScale’: 1, 'removeOutliers’: None, 'resampledPixelSpacing’: None, 'interpolator’: 'sitkBSpline’, 'preCrop’: False, 'padDistance’: 5, 'distances’: [1], 'force2D’: False, 'force2Ddimension’: 0, 'resegmentRange’: None, 'label’: 1, 'additionalInfo’: True, 'binWidth’: 25.0, 'symmetricalGLCM’: True, 'correctMask’: True}
diagnostics	Configuration	EnabledImageTypes	{’Original’: {}}	{’Original’: {}}
diagnostics	Image-original	Hash	533a438effd3030eeb0cea7ca74e9f5887af4945	533a438effd3030eeb0cea7ca74e9f5887af4945
diagnostics	Image-original	Dimensionality	3D	3D
diagnostics	Image-original	Spacing	(3.90625, 3.90625, 4.25)	(3.90625, 3.90625, 4.25)
diagnostics	Image-original	Size	(128, 128, 239)	(128, 128, 239)
diagnostics	Image-original	Mean	22.01189889	22.01189889
diagnostics	Image-original	Minimum	0	0
diagnostics	Image-original	Maximum	2486.789208	2486.789208
diagnostics	Mask-original	Hash	233cb5136e729a3e4f21362aca4475c2ce7026e1	c3c38c72291849f4c9fd08f3f2350db65cd702d3
diagnostics	Mask-original	Spacing	(3.90625, 3.90625, 4.25)	(3.90625, 3.90625, 4.25)
diagnostics	Mask-original	Size	(128, 128, 239)	(128, 128, 239)
diagnostics	Mask-original	BoundingBox	(72, 72, 94, 8, 15, 11)	(39, 71, 93, 10, 12, 10)
diagnostics	Mask-original	VoxelNum	541	443
diagnostics	Mask-original	VolumeNum	1	1
diagnostics	Mask-original	CenterOfMassIndex	(75.80406654343808, 78.79112754158965, 98.68022181146026)	(43.63205417607224, 76.4334085778781, 97.2686230248307)
diagnostics	Mask-original	CenterOfMass	(46.109634935304996, 57.77784195933458, -601.1090573012939)	(-79.56228837471781, 48.56800225733633, -607.1083521444696)
original	shape	Elongation	0.718428907	0.844158396
original	shape	Flatness	0.520675964	0.600962385
original	shape	LeastAxisLength	26.82559939	26.77934631
original	shape	MajorAxisLength	51.52071776	44.56076951
original	shape	Maximum2DDiameterColumn	45.28123922	48.53417771
original	shape	Maximum2DDiameterRow	59.20707339	41.31831464
original	shape	Maximum2DDiameterSlice	59.75413492	48.94517221
original	shape	Maximum3DDiameter	59.90508443	50.13637651
original	shape	MeshVolume	34381.23067	27977.30764
original	shape	MinorAxisLength	37.01397296	37.61634771
original	shape	Sphericity	0.81721643	0.792830881
original	shape	SurfaceArea	6256.930251	5621.35284
original	shape	SurfaceVolumeRatio	0.181986803	0.20092544
original	shape	VoxelVolume	35083.77075	28728.48511
original	firstorder	10Percentile	296.780922	196.64847
original	firstorder	90Percentile	1729.293089	1081.286734
original	firstorder	Energy	550714072.8	173841865.9
original	firstorder	Entropy	5.880349771	5.144364146
original	firstorder	InterquartileRange	826.7166258	456.6205522
original	firstorder	Kurtosis	2.749790628	4.095834657
original	firstorder	Maximum	2486.789208	1871.7509
original	firstorder	MeanAbsoluteDeviation	468.9773551	281.3159101
original	firstorder	Mean	839.3567999	517.6909538
original	firstorder	Median	628.6984404	387.1992456
original	firstorder	Minimum	190.4121598	147.7359567
original	firstorder	Range	2296.377048	1724.014944
original	firstorder	RobustMeanAbsoluteDeviation	352.5233463	195.2900155
original	firstorder	RootMeanSquared	1008.937943	626.4340044
original	firstorder	Skewness	0.940244297	1.300753755
original	firstorder	TotalEnergy	35713726953	11273619540
original	firstorder	Uniformity	0.022348564	0.039179817
original	firstorder	Variance	313435.9349	124415.6382
original	glcm	Autocorrelation	1337.177605	490.8682935
original	glcm	ClusterProminence	7067328.506	1486133.703
original	glcm	ClusterShade	46339.60226	16679.47707
original	glcm	ClusterTendency	1713.395114	672.7994778
original	glcm	Contrast	434.092444	205.4518274
original	glcm	Correlation	0.599616154	0.537346338
original	glcm	DifferenceAverage	15.79439971	10.98590189
original	glcm	DifferenceEntropy	5.25598161	4.770647602
original	glcm	DifferenceVariance	175.1626669	80.70159972
original	glcm	Id	0.143851838	0.186220415
original	glcm	Idm	0.07537084	0.110098784
original	glcm	Idmn	0.956825044	0.962847097
original	glcm	Idn	0.866847358	0.874617615
original	glcm	Imc1	-0.456304457	-0.371594794
original	glcm	Imc2	0.998000784	0.990742836
original	glcm	InverseVariance	0.079726888	0.10876887
original	glcm	JointAverage	31.88255748	19.33349937
original	glcm	JointEnergy	0.001693794	0.002769738
original	glcm	JointEntropy	9.356114223	8.757717212
original	glcm	MCC	0.712991455	0.66080859
original	glcm	MaximumProbability	0.006333344	0.010531668
original	glcm	SumAverage	63.76511495	38.66699874
original	glcm	SumEntropy	6.818176686	6.143977009
original	glcm	SumSquares	536.8718896	219.5628263
original	gldm	DependenceEntropy	6.843322445	6.415485289
original	gldm	DependenceNonUniformity	261.3770795	152.4266366
original	gldm	DependenceNonUniformityNormalized	0.483136931	0.344078186
original	gldm	DependenceVariance	0.634472344	0.853955944
original	gldm	GrayLevelNonUniformity	12.09057301	17.35665914
original	gldm	GrayLevelVariance	501.8990368	198.4882776
original	gldm	HighGrayLevelEmphasis	1235.898336	461.1038375
original	gldm	LargeDependenceEmphasis	2.881700555	4.205417607
original	gldm	LargeDependenceHighGrayLevelEmphasis	2880.072089	1383.89842
original	gldm	LargeDependenceLowGrayLevelEmphasis	0.049963494	0.146062888
original	gldm	LowGrayLevelEmphasis	0.014439637	0.030611278
original	gldm	SmallDependenceEmphasis	0.713459129	0.551884249
original	gldm	SmallDependenceHighGrayLevelEmphasis	941.3063273	300.2885403
original	gldm	SmallDependenceLowGrayLevelEmphasis	0.009915324	0.014809438
original	glrlm	GrayLevelNonUniformity	11.70994981	16.51195929
original	glrlm	GrayLevelNonUniformityNormalized	0.022068484	0.03850109
original	glrlm	GrayLevelVariance	501.3122804	199.2892289
original	glrlm	HighGrayLevelRunEmphasis	1241.486747	465.9229893
original	glrlm	LongRunEmphasis	1.059378422	1.10049669
original	glrlm	LongRunHighGrayLevelEmphasis	1298.7777	497.3345847
original	glrlm	LongRunLowGrayLevelEmphasis	0.015422869	0.034045554
original	glrlm	LowGrayLevelRunEmphasis	0.014378364	0.030441998
original	glrlm	RunEntropy	5.976110053	5.310844322
original	glrlm	RunLengthNonUniformity	510.5775347	402.0003095
original	glrlm	RunLengthNonUniformityNormalized	0.962181014	0.937159747
original	glrlm	RunPercentage	0.980804777	0.968050009
original	glrlm	RunVariance	0.019765585	0.032984177
original	glrlm	ShortRunEmphasis	0.985477316	0.975477745
original	glrlm	ShortRunHighGrayLevelEmphasis	1227.934367	458.0863265
original	glrlm	ShortRunLowGrayLevelEmphasis	0.01412371	0.029570304
original	glszm	GrayLevelNonUniformity	8.225653207	8.724381625
original	glszm	GrayLevelNonUniformityNormalized	0.019538369	0.030828204
original	glszm	GrayLevelVariance	496.6355866	204.9185281
original	glszm	HighGrayLevelZoneEmphasis	1312.477435	529.4275618
original	glszm	LargeAreaEmphasis	2.315914489	3.643109541
original	glszm	LargeAreaHighGrayLevelEmphasis	2311.304038	1293.833922
original	glszm	LargeAreaLowGrayLevelEmphasis	0.0399263	0.134434019
original	glszm	LowGrayLevelZoneEmphasis	0.013797431	0.027308059
original	glszm	SizeZoneNonUniformity	290.2684086	147.3745583
original	glszm	SizeZoneNonUniformityNormalized	0.689473655	0.520758157
original	glszm	SmallAreaEmphasis	0.856105631	0.749508706
original	glszm	SmallAreaHighGrayLevelEmphasis	1148.741358	426.7926749
original	glszm	SmallAreaLowGrayLevelEmphasis	0.011822282	0.019085154
original	glszm	ZoneEntropy	6.607326679	6.315456586
original	glszm	ZonePercentage	0.77818854	0.638826185
original	glszm	ZoneVariance	0.66459792	1.192723096
original	ngtdm	Busyness	0.095294279	0.244693066
original	ngtdm	Coarseness	0.00917816	0.010024537
original	ngtdm	Complexity	30438.04282	9770.860363
original	ngtdm	Contrast	1.363559647	0.788477366
original	ngtdm	Strength	45.03347974	28.76248158

References

Zwanenburg, A., Vallières, M., Abdalah, M. A., Aerts, H. J., Andrearczyk, V., Apte, A., … & Löck, S. (2020). The image biomarker standardization initiative: standardized quantitative radiomics for high-throughput image-based phenotyping. Radiology, 295(2), 328-338., https://doi.org/10.1148/radiol.2020191145

Fedorov A., Beichel R., Kalpathy-Cramer J., Finet J., Fillion-Robin J-C., Pujol S., Bauer C., Jennings D., Fennessy F.M., Sonka M., Buatti J., Aylward S.R., Miller J.V., Pieper S., Kikinis R. 3D Slicer as an Image Computing Platform for the Quantitative Imaging Network. Magnetic Resonance Imaging. 2012 Nov;30(9):1323-41. PMID: 22770690. PMCID: PMC3466397. https://doi.org/10.1016/j.mri.2012.05.001

Kinahan, Paul; Muzi, Mark; Bialecki, Brian; Coombs, Laura. (2017). Data from ACRIN-FLT-Breast. The Cancer Imaging Archive. http://doi.org/10.7937/K9/TCIA.2017.ol20zmxg