Table 7.
Based on the NCT-CRC-HE-100 K and Kather texture 2016 images, a comparison is made of quantitative methods and our proposed model.
| Method | Dataset | Model | Accuracy (%) | Generalizability | Computational complexity |
|---|---|---|---|---|---|
| Ghosh et al.51 | NCT-CRC-HE-100K | Ensemble learning based on the CNN | 96.16 | Medium | High |
| Hamida et al.52 | Kather texture 2016 | ResNet and TL | 96.60 | Medium | High |
| NCT-CRC-HE-100K | ResNet and TL | 99.76 | Medium | High | |
| Kather et al.53 | NCT-CRC-HE-100K | VGG-16 and TL | 98.70 | Medium | Low |
| Chen et al.37 | NCT-CRC-HE-100K | MCAM | 99.68 | Medium | High |
| IL-MCAM | 99.78 | Medium | High | ||
| Alqudah et al.41 | Kaggle | QDA | 97.30 | Medium | Low |
| Kumar et al.43 | NCT-CRC-HE-100K | NCT-CRC-HE-100K | 99.21 | Medium | High |
| The proposed model | Kather texture 2016 | dResNet and DeepSVM | 98.75 | High | Low |
| NCT-CRC-HE-100K | 99.76 | High | Low |