Table 2.

Formulae for radiomic scoring of the distal femur and proximal tibia

Anatomical site	Filters	Feature classes	Variables	Coefficient
The distal femur	–	Intercept	Intercept	160.0959
	Original	Firstorder	Mean	− 0.1090
	Wavelet-H	Glrlm	GrayLevelNonUniformity	1.0751
		Gldm	DependenceNonUniformity	0.4234
	Wavelet-L	Firstorder	Mean	− 0.1037
		Glrlm	RunLengthNonUniformity	0.0001
		Glszm	SizeZoneNonUniformity	0.0365
		Ngtdm	Complexity	0.5758
The proximal tibia	–	Intercept	Intercept	160.0959
	Original	Shape2D	MajorAxisLength	1.8547
			MaximumDiameter	− 2.4652
			Perimeter	1.1009
		Firstorder	10Percentile	− 1.0655
			Range	6.0021
			RobustMeanAbsoluteDeviation	1.4276
			Variance	− 0.0007
		Glcm	JointAverage	− 0.4492
			Glcm_SumAverage	− 1.072093e-13
		Glrlm_	LongRunEmphasis	0.2136
			RunLengthNonUniformity	3.8091
			ShortRunHighGrayLevelEmphasis	7.7804
		Glszm_	LargeAreaEmphasis	0.4598
			SmallAreaHighGrayLevelEmphasis	0.1444
			Complexity	− 4.1813
	Wavelet-H	Firstorder	Variance	− 4.5816
	Wavelet-L		InterquartileRange	0.6352
			Median	− 2.3973
			Minimum	6.3474
		Glcm	ClusterProminence	− 2.8356
		Glrlm	GrayLevelVariance	− 2.3664
			LongRunEmphasis	0.5052
			RunLengthNonUniformity	0.7373
		Glszm	GrayLevelVariance	− 0.0729
			LargeAreaEmphasis	0.2385
			SizeZoneNonUniformity	0.2360
		Ngtdm	Complexity	− 0.0861
		Gldm	DependenceNonUniformity	− 1.0041
			SmallDependenceHighGrayLevelEmphasis	− 0.8489

Formula = Sign(S) × log(∣S∣). S represents the linear combination of the intercept and the sum of the products of coefficients and variables. Sign(S) ensures that the sign of the radiomic score matches the sign of S. Log(∣S∣) is the natural logarithm of the absolute value of S. glrlm gray-level run length matrix, gldm gray-level dependence matrix, glszm gray-level size zone matrix, ngtdm neighborhood gray-tone difference matrix, glcm gray-level co-occurrence matrix