CytoSpectre: a tool for spectral analysis of oriented structures on cellular and subcellular levels

BMC Bioinformatics

Table 5 Performance comparison between CytoSpectre (CS) and FibrilTool (FT) using real images

	CS orientation, error mean ± std (deg)	FT orientation, error mean ± std (deg)	Wilcoxon signed rank test p-value	CS circular variance, Pearson’s r, mean ± std	FT isotropy index, Pearson’s r, mean ± std	Wilcoxon signed rank test p-value
Validation set 1	15.5774 ± 14.5547	21.1937 ± 15.0628	2.2755E-11	0.3018 ± 0.4325	0.3024 ± 0.4251	0.8501
Validation set 2	11.2739 ± 11.9993	41.9884 ± 31.2416	1.0532E-15	0.3574 ± 0.3555	−0.0896 ± 0.1343	0.0122
Validation set 3	15.3819 ± 17.3992	16.0332 ± 14.1959	0.0016	0.7192 ± 0.1505	0.6086 ± 0.1440	7.3242E-04

Absolute errors in degrees (mean ± standard deviation) between mean orientation values estimated by human experts (N = 12 for validation sets 1 and 2, N = 13 for validation set 3) for the images of each validation set (N = 11 for validation set 1, N = 10 for validation sets 2 and 3) and values estimated by CS and FT are shown. Pearson’s linear correlation coefficients (mean ± standard deviation) between anisotropy rankings estimated by human experts (N = 12 for validation sets 1 and 2, N = 13 for validation set 3) and measures of isotropy estimated by CS and FT are also shown for the images of each set (N = 15 for all validation sets). Paired, two-sided Wilcoxon signed rank tests were performed to compare the orientation errors and correlation coefficients of CS and FT and the resulting p-values are given

ISSN: 1471-2105