Table 2. Comparison of different optimizers on the segmentation performance of the proposed model on the LUNA16 dataset.
| Optimizer | DSC | IoU | Precision | Recall |
|---|---|---|---|---|
| Adam (47) | 0.7663 | 0.6401 | 0.7882 | 0.7775 |
| AdamW (48) | 0.7770 | 0.6518 | 0.8255† | 0.7648 |
| SGD (ours) | 0.7921† | 0.6662† | 0.8102 | 0.7993† |
†, the optimal result. Adam, adaptive moment estimation; AdamW, Adam with decoupled weight decay; DSC, Dice similarity coefficient; IoU, intersection over union; LUNA16, Lung Nodule Analysis 2016; SGD, stochastic gradient descent.