TABLE 5.

(A) Three filters, wavelet, LoG, and original, that are most commonly applied to feature families to generate high‐order radiomic features, based on features selected by the feature selection methods in the reviewed studies. A total count indicates the overall use of the filter among selected features. (B) Individual radiomic features and their respective number of appearances for each of the three most commonly used feature families: GLCM, first‐order, and GLSZM. A total count indicates the overall use of the feature family among selected features

A)	Classification		Prognostics
Wavelet	glcm	65	glcm	36
	first‐order	50	first‐order	18
	glszm	37	glszm	8
	gldm	32	glrlm	8
	stats	22	gldm	3
	Total Count	237	Total Count	111
LoG	glcm	26	first‐order	18
	glszm	23	glcm	15
	first‐order	15	glszm	4
	gldm	12	gldm	4
	glrlm	9	glrlm	1
	Total Count	104	Total Count	52
Original	first‐order	11	glcm	11
	glcm	9	shape	7
	glszm	5	first‐order	6
	shape	5	glrlm	5
	glrlm	3	gldm	4
	Total Count	35	Total Count	51

B)	Classification		Prognostics
GLCM	clustershade	16	clustershade	15
	clusterprominence	14	clusterprominence	13
	entropy	14	correlation	7
	correlation	10	inversevariance	7
	imc1	10	correl1	5
	Total Count	173	Total Count	119
First‐order	mean	12	skewness	9
	median	11	energy	6
	skewness	11	10percentile	5
	90percentile	6	maximum	5
	robustmeanabsolutedeviation	6	rms	5
	Total Count	83	Total Count	59
GLSZM	largearealowgraylevelemphasis	8	largeareaemphasis	5
	graylevelnonuniformity	6	graylevelnonuniformitynormalized	4
	zoneentropy	6	graylevelvariance	3
	lowintensitysmallareaemphasis	5	largearealowgraylevelemphasis	3
	sizezonenonuniformitynormalized	5	sizezonenonuniformity	3
	Total Count	76	Total Count	39