Table 2 Ablation experiment on propagation strategies and memory mechanism. Results are presented as mean ± standard deviation from five-fold cross-validation. The best and second-best results are bolded and underlined, respectively.

From: ICH-HPINet: a hybrid propagation interaction network for intelligent and interactive 3D intracerebral hemorrhage segmentation

Method

Private Dataset

Physionet Dataset

Dice\(\uparrow\)

Jaccard\(\uparrow\)

HD\(\downarrow\)

MAE\(\downarrow\)

Dice\(\uparrow\)

Jaccard\(\uparrow\)

HD\(\downarrow\)

MAE\(\downarrow\)

Volume-Pro

0.6554

0.5255

5.96

0.0050

0.6351

0.5152

4.29

0.0180

± 0.026

± 0.023

± 0.51

± 0.0006

± 0.029

± 0.025

± 0.46

± 0.0021

Slice-Pro

0.6468

0.5202

4.26

0.0054

0.4587

0.3728

5.45

0.0148

± 0.027

± 0.024

± 0.39

± 0.0006

± 0.035

± 0.030

± 0.58

± 0.0018

w/o Memory

0.7152

0.5896

4.12

0.0046

0.6598

0.5381

4.85

0.0052

± 0.024

± 0.021

± 0.37

± 0.0005

± 0.028

± 0.024

± 0.49

± 0.0006

ICH-HPINet (Ours)

0.7797

0.6532

3.59

0.0035

0.7467

0.6094

2.96

0.0022

± 0.021

± 0.018

± 0.32

± 0.0004

± 0.024

± 0.021

± 0.28

± 0.0003