Table 1 Comparison of performance of hypothalamic segmentation by four investigated network architectures in the test dataset.
Ground truth | EfficientNetB0 | Inceptionv3 | ResNet50 | VGG16 | |
|---|---|---|---|---|---|
IoU | – | 0.88 ± 0.03 | 0.87 ± 0.02 | 0.87 ± 0.02 | 0.85 ± 0.04 |
Precision | – | 0.87 ± 0.05 | 0.79 ± 0.05 | 0.82 ± 0.05 | 0.82 ± 0.05 |
Recall | – | 0.86 ± 0.05 | 0.93 ± 0.04 | 0.87 ± 0.04 | 0.85 ± 0.11 |
Dice | – | 0.87 ± 0.03 | 0.85 ± 0.03 | 0.85 ± 0.03 | 0.83 ± 0.07 |
Prediction time per image/ms | – | 29 | 31 | 35 | 42 |
Hypothalamic volume/cm3 | 0.82 ± 0.10 | 0.80 ± 0.10 | 0.96 ± 0.12 | 0.87 ± 0.11 | 0.85 ± 0.16 |
p-value | – | 0.336§ | 1.189·10–11§ | 0.0003§ | 0.045$ |