Table 13 Results on the best multicenter models: all presented models use StandardScaling for normalization and SMOTENN for database balancing. MultiC: Multicenter.
Multicenter model | Dataset | Model configuration | Training (%) | Validation | Testing on IDC | Testing on BreaKHis |
|---|---|---|---|---|---|---|
Model IDC | IDC | 2 | 89.46 | 87.87 | 87.55 | 58.16 |
Model BreaKHis | BreaKHis | 2 | 99.27 | 92.34 | 47.20 | 93.28 |
MultiC 1 | IDC ; BreaKHis | 2 | 95.67 | 95.04 | 80.00 | 81.00 |
MultiC 3 | IDC ; BreaKHis | 3 intra-base | 97.24 | 95.83 | 83.67 | 67.33 |
MultiC 4 | IDC ; BreaKHis | 7 | 96.03 | 95.22 | 74.00 | 63.00 |
MultiC 5 | IDC ; BreaKHis | 9 | 99.94 | 99.91 | 77.00 | 69.00 |
MultiC 6 | IDC ; BreaKHis | 10 | 99.98 | 99.94 | 74.43 | 67.56 |
