Table 1 Comparison of the segmentation performances on the local dataset using the CNN–transformer hybrid network containing different modules with the best results in bold
From: A neural network with a human learning paradigm for breast fibroadenoma segmentation in sonography
Methods | DSC ↑ | HD (mm) ↓ | Time (h)↓ | |
|---|---|---|---|---|
Fragmentation module | Aggregation module | |||
NA | NA | 0.869240 | 8.963769 | 4 |
Focus | LogSparse | 0.870811 | 8.026729 | 4 |
C3CBAM | 0.867615 | 7.836283 | 3.75 | |
ProbSparse | 0.869836 | 8.212088 | 3 | |
BottleneckCSP | LogSparse | 0.874429 | 7.846115 | 3.25 |
C3CBAM | 0.855873 | 10.848110 | 2.25 | |
ProbSparse | 0.854521 | 9.704177 | 3.25 | |
C3ECA | LogSparse | 0.875815 | 5.820322 | 2.75 |
C3CBAM | 0.870480 | 8.873851 | 2.25 | |
ProbSparse | 0.863999 | 9.969885 | 2.75 | |