Table 1 Related work highlighting the used datasets, their size, number of classes (C), employed methods, and accuracy
From: WBC image classification and generative models based on convolutional neural network
Study | Dataset | Size | C | Methods | Performance |
|---|---|---|---|---|---|
Wang et al. [19] | Private: hyperspectral blood cell images | N/A | 5 | Morphology, spectral analysis and SVM | 90.00% |
Dorini et al. [20] | CellAtlas | 100 | 5 | Morphological transform. and KNN | 78.51% |
Nazlibilek et al. [27] | Kanbilim dataset [73] | 240 | 5 | Thresholding, ANN and PCA | 95.00% |
Prinyakupt et al. [21] | Private dataset: Rangsit University and | PD: 555 | 5 | Thresholding and NB | PD: 93.70% |
| Â | CellaVision dataset | CV: 2477 | Â | Â | CV: 92.90% |
Abdeldaim et al. [28] | ALL-IDB2 | 260 | 2 | Thresholding, KNN, SVM, NB and DT | KNN: 96.01% |
| Â | Â | Â | Â | Â | SVM: 93.89% |
| Â | Â | Â | Â | Â | NB: 89.97% |
| Â | Â | Â | Â | Â | DT: 86.81% |
Hegde et al. [32] | Private: Kolkata Municipal Corporation | 117 | 5 | Arithmetical operations and ANN | 96.50% |
Ghosh et al. [74] | ALL-IDB | 260 | 2 | CNN | 97.22% |
Rezatofighi et al. [17] | Private: Imam Khomeini Hospital | 400 | 5 | Gram-Schmidt, SVM and ANN | 98.64% |
Habibzadeh et al. [38] | Private [75] | 352 | 4 | CNN | 93.17% |
Liang et al. [76] | BCCD [77] | 364 | 4 | RNN (LSTM) and CNN | 90.79% |
Rawat et al. [39] | Private [78] | 160 | 4 | Ensemble ANN | 95.00% |
Ramesh et al. [36] | Private: University of Utah | 320 | 5 | LDA | 93.90% |
Putzu et al. [79] | ALL-IDB | 260 | 2 | SVM | 92.00% |
Mathur et al. [34] | Private | 237 | 5 | NB | 92.72% |
Ghosh et al. [37] | Private: Kolkata Municipal Corporation | 150 | 5 | Region-based segmentation | N/A |
| Â | Â | Â | Â | Mathematical morphology | N/A |
| Â | Â | Â | Â | Fuzzy logic and RF | N/A |
Su et al. [35] | CellaVision [80] | 450 | 5 | Mathematical morphology | HCNN: 88.89% |
| Â | Â | Â | Â | Hyperrectangular composite NN | SVM: 97.55% |
| Â | Â | Â | Â | SVM and MLP | MLP: 99.1% |
Patil et al. [40] | BCCD [77] | 12,442 | 4 | CNN and RNN | 95.89% |
Toğaçar et al. [41] | BCCD [77] | 12,435 | 4 | AlexNet, GoogLeNet and ResNet | 97.95% |
Mohamed et al. [43] | BCCD [77] | 12,500 | 4 | MobileNet-22 | 97.03% |
Banik et al. [44] | BCCD, ALL-IDB2, JTSC, and CV [80] | 13,371 | 4 | CNN | 94.00% |
Karthikeyan et al. [45] | BCCD [77] | 12,500 | 4 | LSM-TIDC | N/A |
Kutlu et al. [46] | 12,500 | 5 | Regional-based CNN | 97.52% | |
W-Net (this work) | Private: The Catholic University of Korea | 6562 | 5 | CNN | 97% |
W-Net (this work) | LISC public data [17] | 254 | 5 | CNN and further training | 96% |