Table 2. Summary of CT clinical studies with relative correlations with EAT and others adipose tissue compartments.

Authors	AI category	Algorithm and/or software	Adipose tissue compartments	Data	Correlation	Number of patients	Results	Level of evidence*
Kroll et al. [2021] (46)	Deep learning	U-Net 3D	EAT	EAT Volume	Calcium artery score	966	r=0.18	3
			Paracardialadipose tissue	EAT Attenuation			r=−0.09
			Visceral adipose tissue
			Intermuscularadipose tissue
			Subcutaneous adipose tissue
Eisenberg et al. [2020] (47)	Deep learning	Fully automated DL algorithm incorporated into QFAT research software (version 2.0)	EAT	EAT Volume	Major adverse cardiovascular events	2,068	P<0.01	1
				EAT Attenuation			P=0.01
Oikonomouet al. [2019] (48)	Machine learning	Aquarius Workstation V.4.4.11-13, TeraRecon Inc., Foster City, CA, USA for basic segmentation and CaRi-HEART proprietary algorithms	Perivascular adipose tissue	Fat attenuation index	Inflammation	167, 1,575, 44	See Figure 4 for inflammation, vascularity and fibrosis	1, 3, 3
		Caristo Diagnostics Ltd, Oxford, UK		Fat radiomic profile	Vascularity
		Slicer Radiomics extension which incorporates the Pyradiomics library into 3D Slicer (v.4.9.0-2017-12-18 r26813)			Fibrosis
					Major adverse cardiovascular events		P<0.001
					Perivascular changes related to acute myocardial infarction		P<0.001
Lin et al. [2021] (49)	Deep learning	QFAT v2.0, CSMC, Los Angeles, CA, USA	EAT	EAT volume	Major adverse cardiovascular events	2,068	P<0.001	2
			NAFLD	EAT attenuation			P<0.001
				Hepatic attenuation			P=0.003
Tamarappooet al. [2021] (50)	Deep learning	QFAT v2.0, CSMC, Los Angeles, CA, USA	EAT	EAT volume	Long-term risk prediction for cardiac events	1,069	P<0.0001	1
				EAT attenuation			P=0.002

*, SIGN100: Scottish Intercollegiate Guidelines Network 2019. CT, computed tomography; AI, artificial intelligence; EAT, epicardial adipose tissue; NAFLD, non-alcoholic fatty liver disease.