Table 4.
Application of AI in cervical cell classification.
| Reference | Year | Methods | Datasets (Num. of images) | Classes | Results |
|---|---|---|---|---|---|
| Chankong et al. (41) | 2014 | Bayesian classifier KNN ANN | ERUDIT (552) | 4-class | Accuracy 96.20% |
| 2-class | Accuracy 97.83% | ||||
| Herlev (917) | 7-class | Accuracy 93.78% | |||
| 2-class | Accuracy 99.27% | ||||
| LCH (300) | 4-class | Accuracy 95.00% | |||
| 2-class | Accuracy 97.00% | ||||
| Borakden et al. (53) | 2017 | Ensemble classifier: LSSVM MLP RF | Cell level (1610) | 2-class | Accuracy 99.07% |
| Specificity 98.90% | |||||
| Smear level (1320) | 3-class | Accuracy 98.11% | |||
| Specificity 99.35% | |||||
| Hervel (917) | 2-class | Accuracy 96.51% | |||
| Specificity 89.67% | |||||
| Zhang et al. (54) | 2017 | CNN; Transfer learning | Herlev (917) | 7-class | Accuracy 98.30% |
| Specificity 98.30% | |||||
| HEMLBC (2370) | 2-class | Accuracy 98.60% | |||
| Specificity 99.00% | |||||
| sensitivity 98.30% | |||||
| Hussain et al. (52) | 2020 | CNN; Transfer learning | LBC (own) (1670), Conventional(own) (1320) | 4-class | Accuracy 98.90% |
| Sensitivity 79.80% | |||||
| Specificity 97.90% | |||||
| Shi J et al. (55) | 2020 | CGN | SIPAKMeD (4049) | 5-class | Accuracy 98.37% |
| Sensitivity 99.80% | |||||
| MOTIC (25378) | 7-class | Accuracy 94.93% | |||
| Sensitivity 92.98% | |||||
| Rahaman et al. (56) | 2021 | HDFF | Herlev (917) | 2-class | Accuracy 98.32% |
| 7-class | Accuracy 90.32% | ||||
| SIPAKMeD (4049) | 2-class | Accuracy 90.32% | |||
| 5-class | Accuracy 99.14% |
KNN, K- Nearest Neighbor; ANN, Artificial Neural Network; LSSVM, Least Squares Support Vector Machine.
CNN, convolutional neural network; CGN, graph convolution network; HDFF, hybrid deep feature fusion techniques.