Skip to main content

Table 2 Mean Dice similarity coefficient (DSC) for the lumen segmentations after registration

From: Regional registration of whole slide image stacks containing major histological artifacts

Method

DSC

Training time

Exec. time

Moles Lopez et al. [5]—1 Round

\(0.74 \pm 0.19\)

5.8 mina

Moles Lopez et al. [5]—2 Rounds

\(0.81 \pm 0.15\)

6.4mina

Wang and Chen [6]—1 Round

\(0.82 \pm 0.12\)

2.8 min b

Wang and Chen [6]—2 Rounds

\(0.77 \pm 0.22\)

3.0 minb

Balakrishnan et al. [7]—Patch size 256

\(0.79 \pm 0.16\)

80.5 mina

0.34 mina

Balakrishnan et al. [7]—Patch size 512

\(0.77 \pm 0.16\)

316.9 mina

0.35 mina

Proposed Algorithm

\(0.84 \pm 0.11\)

3.4 mina

Proposed Algorithm followed by Moles Lopez et al. [5]

\(\mathbf {0.86 \pm 0.08}\)

5.6 mina

  1. The DSC was measured for lumen segmentations of 20 blood vessels for 5 consecutive slices after registration using the method proposed by Moles Lopez et al. [5] (Moles Lopez et al. [5] 1 Round), the regional version of this method (Moles Lopez et al. [5] 2 Rounds), the method proposed by Wang and Chen [6] (Wang and Chen [6]—1 Round), the regional version of this method (Wang and Chen [6]—2 Rounds), and the patch-based method proposed by Balakrishnan et al. [7] with patch sizes 256 \(\times\) 256 pixels and 512 \(\times\) 512 pixels. Their performance was compared with those of the proposed algorithm, and the proposed method followed by fine registration using Moles Lopez et al. [5]. The time required for training and executing the algorithms on 5 consecutive image slices is presented in the third and fourth columns
  2. aUbuntu 19.04.4 LTS 64-bit, Intel Core i7-6700 CPU 3.40 GHz × 8, 31.4GB RAM
  3. bWindows 8.1 Pro 64-bit, Intel Core i7-4720HQ CPU 2.60GHz, 11.9GB RAM