Table 7 Ablation study results showing the impact of different components (RDMS, Transform, and HS-CBAM-FPN) on model performance across two datasets.
From: Multi scale deep learning quantifies Ki67 index in breast cancer histopathology images
Dataset | RDMS | Transform | HS-CBAM-FPN | Avg. F1 (%) |
|---|---|---|---|---|
SHIDC-BC-Ki-67 | \(\times\) | \(\times\) | \(\times\) | 79.89 |
\(\checkmark\) | \(\times\) | \(\times\) | 80.24 | |
\(\times\) | \(\checkmark\) | \(\times\) | 81.73 | |
\(\times\) | \(\times\) | \(\checkmark\) | 81.45 | |
\(\checkmark\) | \(\checkmark\) | \(\times\) | 82.89 | |
\(\checkmark\) | \(\times\) | \(\checkmark\) | 82.41 | |
\(\times\) | \(\checkmark\) | \(\checkmark\) | 84.45 | |
\(\checkmark\) | \(\checkmark\) | \(\checkmark\) | 85.79 | |
BCData | \(\times\) | \(\times\) | \(\times\) | 79.25 |
\(\checkmark\) | \(\times\) | \(\times\) | 79.89 | |
\(\times\) | \(\checkmark\) | \(\times\) | 81.05 | |
\(\times\) | \(\times\) | \(\checkmark\) | 81.23 | |
\(\checkmark\) | \(\checkmark\) | \(\times\) | 82.71 | |
\(\checkmark\) | \(\times\) | \(\checkmark\) | 83.12 | |
\(\times\) | \(\checkmark\) | \(\checkmark\) | 83.79 | |
\(\checkmark\) | \(\checkmark\) | \(\checkmark\) | 84.25 |
