Table 3.
The results for the 8 considered methods on the stroke and white matter hyperintensity (SWMH) data set. The best results are shown in bold.
| Methods | Ischemic stroke segmentation | WMHs segmentation | ||
|---|---|---|---|---|
| DC | HD | DC | HD | |
| U-Net (Ronneberger et al., 2015) | 52.35 (±7.50) | 6.02 (±5.32) | 50.06 (±8.18) | 7.06 (±6.01) |
| uResNet (Guerrero et al., 2017) | 70.80 (±9.90) | 3.25 (±1.92) | 67.16 (±7.20) | 2.97 (±1.97) |
| RA-UNet (Jin et al., 2018) | 72.95 (±7.20) | 3.16 (±1.99) | 71.76 (±6.50) | 2.67 (±1.28) |
| FC-ResNet (Drozdzal et al., 2017) | 73.50 (±7.60) | 3.08 (±1.94) | 71.20 (±7.80) | 2.63 (±1.38) |
| MRFNet (Liu et al., 2020d) | 77.04 (±2.35) | 2.94 (±1.31) | 73.65 (±3.38) | 2.47 (±1.04) |
| MedT (Valanarasu et al., 2021) | 79.00 (±2.99) | 3.01 (±1.20) | 77.98 (±2.01) | 2.48 (±1.10) |
| TransUNet (Chen et al., 2021) | 79.06 (±2.76) | 2.79 (±0.99) | 78.02 (±3.21) | 2.38 (±1.99) |
| LLRHNet | 79.10 (±2.63) | 2.70 (±1.51) | 78.02 (±3.10) | 2.27 (±2.01) |