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

Table 5 Mean ± std for nuclear segmentation evaluation metrics using trained SplineDist and UF-UNet models

Metric	Type	Pre-training		Post-training
Metric	Type	SplineDist	UF-UNet	SplineDist	UF-UNet
Region based measurements
F1 score	ROI	0.81 ± 0.09	0.16 ± 0.20	0.89 ± 0.05	0.80 ± 0.09
IoU score	ROI	0.69 ± 0.12	0.11 ± 0.17	0.81 ± 0.07	0.68 ± 0.13
False discovery rate	ROI	0.14 ± 0.09	0.72 ± 0.21	0.06 ± 0.05	0.18 ± 0.10
Fowlkes–Mallows index	ROI	0.81 ± 0.09	0.18 ± 0.19	0.89 ± 0.05	0.80 ± 0.09
Pixel based measurements
IoU score	Pixel	0.72 ± 0.11	0.45 ± 0.20	0.87 ± 0.03	0.69 ± 0.13
Cohen’s kappa index	Pixel	0.09 ± 0.08	0.39 ± 0.18	0.88 ± 0.04	0.72 ± 0.16
Matthews correlation coefficient	Pixel	0.75 ± 0.12	0.37 ± 0.25	0.89 ± 0.04	0.75 ± 0.13
False discovery rate	Pixel	0.76 ± 0.10	0.39 ± 0.25	0.06 ± 0.02	0.02 ± 0.03
Feature based measurements
Intersection area	Feature	0.82 ± 0.07	0.37 ± 0.22	0.87 ± 0.07	0.79 ± 0.10
Intersection perimeter	Feature	0.77 ± 0.10	0.32 ± 0.20	0.82 ± 0.09	0.73 ± 0.11
Intersection mean intensity	Feature	0.82 ± 0.08	0.52 ± 0.16	0.89 ± 0.05	0.80 ± 0.08
Intersection solidity	Feature	0.30 ± 0.20	0.24 ± 0.17	0.35 ± 0.22	0.66 ± 0.12

ISSN: 1471-2105