Table 7.
Comparison of deep learning methods for segmenting coarse-grained brain structures on the SchizBull 2008 testing dataset.
Methods | CNNs | Parameters | Dice Score (mean ± standard deviation) |
---|---|---|---|
U-Net (Çiçek et al., 2016) | 3D | 6.444 × 106 | 0.857±0.097 |
SD-Net (Roy et al., 2017) | 2D | 5.7 × 105 | 0.856±0.098 |
U-Net (Ronneberger et al., 2015) | 2D | 5.178 × 106 | 0.862±0.096 |
QuickNAT V2 (Roy et al., 2018) | 2D | 3.551 × 106 | 0.862±0.095 |
ACEnet | 2D | 4.142 × 106 | 0.881±0.074 |