Skip to main content

Table 2 PQ and JI scores on the independent datasets

From: Marker-controlled watershed with deep edge emphasis and optimized H-minima transform for automatic segmentation of densely cultivated 3D cell nuclei

Method

Neurosphere

Embryo

Liver

hiPSC

 

PQ

JI

PQ

JI

PQ

PQ

IF3DImageJSuite [40, 55]

0.02

0.23

0.64

0.65

NA

NA

LoS [40, 56]

0.20

0.40

0.40

0.51

NA

NA

MINS [40, 57]

0.51

0.56

0.79

0.79

NA

NA

OpenSegSPIM [40, 58]

0.58

0.61

0.36

0.48

NA

NA

RACE [40, 59]

0.03

0.39

0.00

0.15

NA

NA

SAMA [40, 60]

0.00

0.12

0.40

0.49

NA

NA

Vaa3D [40, 61]

0.45

0.60

0.40

0.52

NA

NA

XPIWIT [40, 62]

0.59

0.62

0.73

0.74

NA

NA

3D-Cell-Annotator [40, 41]

0.64

0.69

0.80

0.80

NA

NA

\(\text {U-Net-Cell}*\)

0.51 ± 0.03

0.58 ± 0.02

0.64 ± 0.03

0.68 ± 0.02

0.67 ± 0.02

0.59 ± 0.04

\(\text {U-Nets+SWS}*\)

0.65 ± 0.01

0.68 ± 0.01

0.76 ± 0.03

0.77 ± 0.02

0.71 ± 0.01

0.75 ± 0.04

\(B,M_{3DE},-,*\)

0.66 ± 0.01

0.70 ± 0.01

0.78 ± 0.01

0.78 ± 0.01

0.71 ± 0.01

0.70 ± 0.16

\(C,M_{3DE},S,*\)

0.67 ± 0.01

0.70 ± 0.01

0.78 ± 0.01

0.78 ± 0.01

0.71 ± 0.01

0.69 ± 0.14

\(C,M_{2DE}, S,*\)

0.66 ± 0.01

0.69 ± 0.01

0.80 ± 0.02

0.81 ± 0.02

0.71 ± 0.01

0.74 ± 0.03

  1. The configurations are specified by \(\{w,M,S,*\}\): w denotes watershed method, \(M \in \{M_{3D},M_{3DE},M_{2DE}\}\), which denotes 3D mask, 3D and 2D edge emphasizing mask, respectively, and \(S \in \{S,-\}\), which refers to use or exclusion of seeds. We also report PQ and JI values of the reference software computed from the ground truth and result files provided by Piccinini et al. [40] for the Neurosphere and Embryo datasets. The configurations, U-Net+SWS* and U-Net-Cell* both had twelve different settings: the average scores and their standard deviations of these settings are illustrated. PQ = Panoptic Quality, JI = Jaccard Index. The best value for each dataset and evaluation metric combination is bolded