Maximizing Kolmogorov Complexity for accurate and robust bright field cell segmentation

BMC Bioinformatics

Table 3 Segmentation results for three data sets for human embryonic kidney cells (HEK 293T)

Data set one	Maximal-information	SephaCe	Correlation	t- stat	P(T≤ t)one-tail
Correct segmentation t p	9.12	5.76	0.3970	9.4557	0.0
Unexpected areas f p	0.68	0.80	0.2355	-0.5492	0.2939
Missing cells f n	1.60	4.72	-0.0909	-9.0929	0.0
Precision Pr	93.20%	89.36%	0.3295	1.4461	0.0805
Recall Re	85.37%	54.34%	-0.2903	8.2830	0.0
Data set Two	*Maximal-information*	*SephaCe*	Correlation	t stat	P ( T ≤ t )one-tail
Correct segmentation t p	13.35	12.60	0.4344	3.4701	0.0012
Unexpected areas f p	1.15	2.20	0.1633	-4.0977	0.0003
Missing cells f n	0.50	1.25	0.2939	-3.4701	0.0012
Precision Pr	92.30%	85.45%	0.1690	4.3714	0.0001
Recall Re	96.40 %	91.08%	0.2822	3.4407	0.0013
Data set three	*Maximal-information*	*SephaCe*	Correlation	t stat	P ( T ≤ t )one-tail
Correct segmentation t p	15.56	11.86	0.4549	10.18	0.0
Unexpected areas f p	1.72	2.00	0.3642	-0.9434	0.1759
Missing cells f n	2.81	6.36	0.4926	-9.9501	0.0
Precision Pr	91.66%	86.23%	0.3887	2.6898	0.0
Recall Re	85.94%	65.21%	0.4256	10.12	0.0

tp is the average count of correctly identified cells, fp is unexpected segmentations and fn is cells that were missed. Recall and precision are given as percentages.

ISSN: 1471-2105