Research Article
A Neural-Network-Based Approach to White Blood Cell Classification
Table 4
The comparisons of the classification rates among different classification systems.
| Method | Number of types | Segmentation | Classifier | Overall rate | Number of images |
| Ours | 5 | Discriminating region | MLP | 99.11% | 450 | Ours | 5 | Discriminating region | SVM | 97.55% | 450 | Ours | 5 | Discriminating region | HRCNN | 88.89% | 450 | Rezatofighi et al. [1] | 5 | Gram-Schmidt orthogonalization and snake | SVM | 86.10% | 400 | Tabrizi et al. [2] | 5 | Gram-Schmidt orthogonalization and snake | LVQ | 94.10% | 400 | Ghosh et al. [3] | 5 | Watershed | Bayes classifier | 83.2% | 150 | Young [4] | 5 | Histogram threshold | Distance classifier | 92.46% | 199 | Yampri et al. [5] | 5 | Automatic thresholding and adaptive contour | Minimized error | 96.0% | 100 |
Bikhet et al. [6] | 5 | Entropy threshold and iterative threshold | Distance classifier | 90.14% | 71 |
Piuri and Scotti [7] | 5 | Opening and Canny edge detector | KNN, FF-NN, and RBF | 92%~82% | 113 |
|
|