. Author manuscript; available in PMC: 2018 Aug 5.

Published in final edited form as: IEEE Trans Med Imaging. 2018 Feb 14;37(8):1822–1834. doi: 10.1109/TMI.2018.2806309

Table V.

Mean Dice scores for previous abdominal CT multi-organ segmentation methods. Different data sets and segmentation of unlisted organs preclude direct comparisons.

Reference	Method	Mean Dice scores %; higher is better
		Spl.	L.Kid.	Gall.	Esoph.	Liv.	Stom.	Panc.	Duod.
Cerrolaza [5]	SM	0.86	0.69	0.83		0.82	0.87	0.74
Heinrich [34] [50]	MA	0.92	0.92	0.60	0.69	0.95	0.80	0.74
Hu [19]	DL⁺	0.94	0.95			0.96
Kechichian [51]	MA⁺	0.84	0.86	0.14		0.93		0.35
Larsson [20]	MADL	0.93	0.91	0.62	0.66	0.95	0.78	0.60
Oda [52]	MA⁺	0.88^††	0.90^††			0.94^††		0.62^††
Okada [6]	SM⁺	0.93	0.94	0.67				0.73
Roth [31]	DL	0.91		0.71		0.93	0.84	0.63
Shimizu [10]	MA	0.91^†	0.88^†	0.77^†	0.37^†	0.94^†	0.55^†	0.53^†
Suzuki [9]	MA	0.88^†	0.65^†	0.25^†		0.85^†	0.07^†	0.46^†
Tong [8]	MA	0.92	0.93			0.95		0.70
Wang [53]	MA	0.93	0.92			0.95		0.66
Xu [7]	MA	0.90	0.84	0.27	0.43	0.91	0.55	0.45
Zhou [18]	DL	0.92^†	0.91^†	0.65^†	0.43^†	0.95^†	0.60^†^‡	0.62^†	0.60^†^‡
Zografos [54]	RF	0.92^†	0.92^†			0.91^†		0.59^†

DEEDS+JLF	MA	0.87	0.90	0.60	0.64	0.93	0.81	0.72	0.56
VoxResNet	DL	0.90	0.89	0.69	0.65	0.94	0.84	0.72	0.60
VNet	DL	0.92	0.90	0.64	0.65	0.94	0.81	0.67	0.56
Proposed	DL	0.95	0.93	0.73	0.71	0.95	0.87	0.75	0.63

MA=multi-atlas label fusion; SM=statistical shape/appearance model; RF=registration-free; DL=deep learning.

⁺

denotes processing with methods commonly used in registration-free methods.

^†

estimated from mean Jaccard index;

^††

estimated from mean precision and recall.

^‡

Stomach and duodenum were segmented as one class.