Table 4 Some model training parameters changed using the CBIS-DDSM and Duke-Breast-Cancer-MRI datasets, and the Reduce On Plateau parameters for the Learning Rate (LR) refinement.
From: The impact of pre-processing techniques on deep learning breast image segmentation
Dataset | Initial LR | LR Factor | LR Patience | LR Threshold | Channels | Strides |
|---|---|---|---|---|---|---|
CBIS-DDSM | 0.0001 | 0.7 | 100 | 0.0001 | (16, 32, 64, 128) | (2, 2, 2) |
Duke-Breast-Cancer-MRI | 0.001 | 0.7 | 100 | 0.0001 | (16, 32, 64, 128) | (2, 2, 2) |
