Table 2.

Cross-sectional area segmentation using the ground truth-derived and DLM-derived levels.

Parameter	Internal validation dataset			External validation dataset
	SMA	Sfat	Vfat	SMA	Sfat	Vfat
All subjects (n = 1082)
CSA from GT (cm2)	140.88 ± 34.53	140.90 ± 56.71	114.53 ± 65.05	132.76 ± 31.25	133.15 ± 62.16	110.59 ± 64.29
CSA from DLM (cm2)	140.53 ± 34.20	141.98 ± 56.60	115.93 ± 65.40	130.07 ± 31.07	135.54 ± 62.64	110.72 ± 65.19
p value*	0.874	0.764	0.736	0.139	0.492	0.973
CSA error (%)	1.38 ± 1.46	3.51 ± 5.41	4.00 ± 6.35	3.10 ± 2.85	4.54 ± 6.34	4.26 ± 6.47
Subjects with technical success (n = 1004)
CSA from GT (cm2)	141.20 ± 34.46	138.85 ± 55.86	112.42 ± 64.73	132.75 ± 31.15	133.99 ± 62.82	110.88 ± 64.18
CSA from DLM (cm2)	140.87 ± 34.06	140.47 ± 55.72	114.11 ± 64.95	130.14 ± 31.00	136.73 ± 63.15	111.13 ± 65.06
p value*	0.883	0.659	0.692	0.167	0.474	0.950
CSA error (%)	1.22 ± 1.08	2.31 ± 2.21	2.97 ± 3.21	2.86 ± 2.57	3.39 ± 2.78	3.36 ± 4.68
Subjects with technical failure (n = 78)
CSA from GT (cm2)	136.33 ± 35.18	169.78 ± 60.54	144.23 ± 62.27	132.97 ± 32.3	123.03 ± 52.30	107.10 ± 65.61
CSA from DLM (cm2)	135.77 ± 35.80	163.24 ± 64.41	141.53 ± 66.37	129.21 ± 31.93	122.66 ± 54.42	105.78 ± 66.50
p value*	0.949	0.672	0.865	0.579	0.974	0.924
CSA error (%)	3.68 ± 3.19	20.42 ± 8.14	18.37 ± 15.68	6.01 ± 4.18	18.28 ± 15.16	15.06 ± 12.56
p value§	< 0.01	< 0.01	< 0.01	< 0.01	< 0.01	< 0.01

Data are presented as mean ± standard deviation.

*The p-value is calculated from Student t-test comparing the GT CSA and the CSA determined using the DLM.

^§The p-value is calculated from Student t-test comparing CSA errors between subjects with technical success and subjects with technical failure.

CSA cross-sectional area, DLM deep learning model, GT ground truth, Sfat subcutaneous fat area, SMA skeletal muscle area, Vfat visceral fat area.