Benchmark for multi-cellular segmentation of bright field microscopy images

BMC Bioinformatics

Table 1 Evaluation of the three designated tools on the eight available datasets

Algorithm/Dataset	Init	SN15	Melanoma	TScratch	Scatter	Microfluidics	HEK293	MDCK
Mean F-Measure	(N = 28)	(N = 54)	(N = 20)	(N = 24)	(N = 6)	(N = 13)	(N = 12)	(N = 14)
(Median F-Measure)
[F-Measure Adjusted]
Tscratch (Geback et al. 2009)	0.96	0.96	0.88	0.94	0.47	0.42	0.90	0.92
Tscratch (Geback et al. 2009)	(0.96)	(0.97)	(0.90)	(0.93)	(0.47)	(0.41)	(0.91)	(0.93)
MultiCellSeg (Zaritsky et al. 2011)	0.98	0.97	0.85	0.93	0.55	0.35	0.95	0.96
MultiCellSeg (Zaritsky et al. 2011)	(0.98)	(0.98)	(0.91)	(0.95)	(0.56)	(0.45)	(0.95)	(0.98)
Topman et al. 2011	0.98	0.95	0.93	0.78	0.58	0.63	0.85	0.89
	(0.98)	(0.97)	(0.93)	(0.76)	(0.60)	(0.63)	(0.87)	(0.93)
	[0.97]	[0.96]	[0.93]	[0.84]	[0.52]	[0.61]	[0.84]	[0.93]

F-measure was used for evaluation in three forms: mean F-measure of images in the dataset, median, and mean after threshold adjustment on the training set (for [10]).
Best mean F-measure performance is marked in bold.

ISSN: 1471-2105