Unbiased image segmentation assessment toolkit for quantitative differentiation of state-of-the-art algorithms and pipelines

Table 3 Mean ± std for pixel-based metrics using default models for nuclear and cytoplasm segmentations

Method	Region	IoU score	False discovery rate	Cohen’s kappa index	Matthews corr. coefficient
Mesmer	Nucleus	0.88 ± 0.10	0.08 ± 0.11	0.89 ± 0.10	0.90 ± 0.09
CellPose	Nucleus	0.69 ± 0.18	0.02 ± 0.02	0.73 ± 0.19	0.75 ± 0.17
SplineDist	Nucleus	0.72 ± 0.11	0.09 ± 0.08	0.75 ± 0.12	0.76 ± 0.10
Columbus	Nucleus	0.70 ± 0.14	0.07 ± 0.07	0.74 ± 0.13	0.76 ± 0.11
AICS	Nucleus	0.67 ± 0.09	0.02 ± 0.04	0.71 ± 0.13	0.74 ± 0.10
UF-UNet	Nucleus	0.45 ± 0.20	0.39 ± 0.18	0.37 ± 0.25	0.39 ± 0.25
Mesmer	Cytoplasm	0.83 ± 0.09	0.08 ± 0.07	0.71 ± 0.17	0.73 ± 0.16
CellPose	Cytoplasm	0.33 ± 0.19	0.63 ± 0.21	0.20 ± 0.17	0.25 ± 0.17
SplineDist	Cytoplasm	N/A	N/A	N/A	N/A
Columbus	Cytoplasm	0.62 ± 0.17	0.19 ± 18	0.35 ± 0.21	0.38 ± 0.21
AICS	Cytoplasm	0.17 ± 0.17	0.81 ± 0.18	0.06 ± 0.09	0.11 ± 0.13
UF-UNet	Cytoplasm	N/A	N/A	N/A	N/A

ISSN: 1471-2105