Table 8 Ablation study on the RLKA kernel size (k).

From: SwinCLNet: a robust framework for brain tumor segmentation via shifted window attention and cross-scale fusion

Kernel size (k)

Mean dice (%) \(\uparrow\)

Parameters (M) \(\downarrow\)

FLOPs (G) \(\downarrow\)

Inference (ms) \(\downarrow\)

3

88.45 ± 0.29

31.76

147.1

172

5

88.50 ± 0.28

31.84

149.1

173

7

88.54 ± 0.27

31.92

151.0

175

9 (Proposed)

88.56 ± 0.24

32.00

153.0

177

11

88.56 ± 0.21

32.08

155.0

180