Table 1.

Overview of the image comparison. Image comparison calculated as mean ($\pm 1\sigma$) and range ([min;max]) of the test dataset (30 patients; 10 patients for each treatment site) compared to the reference dataset in terms of mean absolute error (MAE) and mean error (ME) between the test (Test) image minus the reference (Ref) image.

Site		Head-and-Neck		Lung		Breast
Test	Ref	MAE	ME	MAE	ME	MAE	ME
		[HU]	[HU]	[HU]	[HU]	[HU]	[HU]
CBCT	rCT	195 ± 20 [160;230]	−122 ± 33 [−183;−71]	219 ± 44 [133;280]	153 ± 48 [94;230]	152 ± 40 [98;213]	71 ± 37[7;115]
sCT single networka	rCT	53 ± 12 [37;77]	−3 ± 7 [−15;10]	83 ± 10 [72;104]	−2 ± 11 [−25;10]	66 ± 18 [41;95]	−6 ± 13[−24;13]
sCT separate networksb	rCT	51 ± 12 [35;74]	−6 ± 6 [−16;4]	86 ± 9 [73;105]	−5 ± 14 [−28;10]	67 ± 18 [41;98]	−5 ± 11[−18;14]
CT	rCT	63 ± 17 [−40;90]	−18 ± 15 [−46;3]	94 ± 23 [68;146]	9 ± 22 [−33;36]	63 ± 24 [40;115]	8 ± 20[−14;54]

^a sCT obtained from a single network trained on all the anatomical sites.

^b sCT obtained from three different networks trained on each anatomical site.