Table 4.
Segmentation performance (mean ± standard deviation of Dice Score) of our methods with different training strategies.
Datasets | First Stage | End-to-End | Two Stages |
---|---|---|---|
2012 MALC (27 structures) | 0.885±0.065 | 0.885±0.065 | 0.891±0.057 |
SchizBull 2008 | 0.872±0.090 | 0.872±0.089 | 0.881±0.074 |
2012 MALC (133 structures) | 0.737±0.164 | 0.746±0.143 | 0.762±0.136 |
Mindboggle-101 | 0.815±0.075 | 0.820±0.076 | 0.825±0.074 |