Table 5 Comparison of descriptive statistics from models trained with segments, segments with vessel (color channel) and segments with spleen on validation set (C_RTVal).

From: Fully automated deep learning based auto-contouring of liver segments and spleen on contrast-enhanced CT images

	Seg 1,			Seg 2,			Seg 3,			Seg 4,			Seg 5–8,			Spleen, N = 40¹
	N = 40			N = 40¹			N = 40¹			N = 40¹			N = 40¹
	(IntraMD^2a = 0.88)			(IntraMD^2a = 0.88)			(IntraMD^2a = 0.94)			(IntraMD^2a = 0.92)			(IntraMD^2a = 0.99)
	(InterMD^2b = 0.82)			(InterMD^2b = 0.85)			(InterMD^2b = 0.91)			(InterMD^2b = 0.88)			(InterMD^2b = 0.96)
	M_nnUnet¹	M_seg+spleen¹	M_vess¹	M_nnUnet¹	M_seg+spleen¹	M_vess¹	M_nnUnet¹	M_seg+spleen¹	M_vess¹	M_nnUnet¹	M_seg+spleen¹	M_vess¹	M_nnUnet¹	M_seg+spleen¹	M_vess¹	M_seg+spleen¹
Dice	0.93	0.93	0.93	0.86	0.86	0.84	0.91	0.91	0.91	0.91	0.90	0.89	0.97	0.97	0.97	0.99
	0.90, 0.07	0.90, 0.06	0.90, 0.07	0.83, 0.09	0.83, 0.10	0.82, 0.10	0.89, 0.10	0.89, 0.10	0.88, 0.12	0.88, 0.08	0.88, 0.08	0.87, 0.09	0.97, 0.03	0.97, 0.02	0.96, 0.03	0.99, 0.01
	0.98, 0.70	0.98, 0.75	0.97, 0.73	0.94, 0.51	0.94, 0.48	0.93, 0.47	0.96, 0.36	0.96, 0.36	0.96, 0.21	0.97, 0.56	0.97, 0.55	0.97, 0.57	0.99, 0.86	0.99, 0.91	0.99, 0.84	1.00, 0.96
\({\text{HD}}_{95}^{3}\)	3.1	3.1	3	8.0	7.5	8	6.6	6.5	7	6.5	6.2	7	4.5	4.4	6	0.9
	4.6, 6.1	4.6, 5.7	5, 6	9.5, 6.0	8.9, 5.1	9, 5	8.0, 5.5	8.4, 6.7	11, 13	8.3, 6.7	8.4, 6.7	10, 8	7.6, 8.8	7.0, 7.5	9, 12	0.9, 0.7
	37.1, 0.8	34.4, 0.8	34, 1	33.4, 3.1	25.8, 2.7	30, 3	28.9, 2.5	32.3, 2.5	81, 2	36.7, 2.1	36.5, 1.5	41, 2	47.0, 0.7	38.2, 0.7	54, 1	2.7, 0.0
Avg. \({\text{HD}}_{{\text{A}}}^{3}\)	0.14	0.13	0.15	0.48	0.45	0.57	0.31	0.28	0.32	0.24	0.24	0.33	0.06	0.06	0.08	0.01
	0.30, 0.61	0.28, 0.50	0.29, 0.50	0.69, 0.62	0.66, 0.60	0.70, 0.60	0.49, 0.87	0.52, 0.97	0.64, 1.33	0.47, 0.55	0.47, 0.52	0.64, 0.90	0.18, 0.29	0.13, 0.19	0.27, 0.61	0.02, 0.01
	3.85, 0.02	3.14, 0.02	3.16, 0.03	2.92, 0.11	3.16, 0.15	3.02, 0.16	5.37, 0.08	5.54, 0.08	7.97, 0.09	2.71, 0.04	2.33, 0.04	4.57, 0.05	1.54, 0.01	1.03, 0.01	3.38, 0.01	0.06, 0.00
PDV	7	5	6	11	10	10	5	5	5	7	6	8	2	2	3	0
	8, 10	8, 10	9, 12	21, 35	21, 36	23, 40	8, 8	8, 8	9, 9	8, 7	9, 7	11, 10	4, 4	3, 3	4, 4	1, 1
	45, 0	45, 0	63, 0	173, 0	177, 0	218, 0	27, 0	29, 0	30, 0	30, 0	25, 1	52, 1	22, 0	12, 0	18, 0	4, 0

¹Data in each cell is organized as row 1 = Median, row 2 = Mean, Standard deviation, row 3 = Max, Min; ¹M_nnUnet = model trained with segments only, M_seg+spleen = Model trained with Segments and Spleen as labels, M_vess = Model trained with segments as label and vessel as color channel.
^2aIntraMD = Intra-observer mean dice.
^2bInterMD = Inter-observer mean dice.
³HD = Hausdorff distance (95 = 95th percentile and A = Average in mm); Wilcoxon signed rank test with Bonferroni adjustment showed p > 0.05 in M_nnU-Net vs. M_seg+spleen for all. In M_seg+spleen vs. M_vess, segment 3 showed p < 0.05 in DSC and HD_A. In M_nnU-Net vs. M_vess, segment 3 showed p < 0.05 in HD₉₅ and segment 4 showed p < 0.05 in PDV and HD₉₅.

Back to article page

Table 5 Comparison of descriptive statistics from models trained with segments, segments with vessel (color channel) and segments with spleen on validation set (C_RTVal).

Search

Quick links