Table 7.
Different methods for detection of hemorrhages.
| Algorithm | Image processing techniques | Database | Color space | Sensitivity | Specificity | Accuracy |
|---|---|---|---|---|---|---|
| Hatanaka et al. [70] | Gamma correction, density analysis | Private data | HSV | 80% | 88% | — |
|
| ||||||
| Tang et al. [71] | Splat feature, filter approach, and wrapper approach | Messidor | Gray scale | — | — | 0.96 AUC at splat level and 0.87 at image Level |
|
| ||||||
| Saleh and Eswaran [72] | H-maxima transformation and multilevel Thresholding |
Private data | Green channel | 84.31% and 87.53% | 93.63% and 95.08% | — |
|
| ||||||
| Lachure et al. [73] | SVM, KNN classifier | Messidor, DB-rect | HIS | 90% | 100% | — |
|
| ||||||
| Sudha and Thirupurasundari [74] | Median filter, adaptive histogram, and KNN Classifier |
Messidor | Gray scale | 100% | 90% | — |
|
| ||||||
| Sharma et al. [75] | Dynamic thresholding | DIARETDB1 | Green channel | 90% | — | — |
|
| ||||||
| Akram et al. [76] | Gaussian Mixture Model, Filter Bank | DRIVE, STARE DIARETDB, Messidor |
Green channel | 97.83% | 98.36% | 98.12% |
|
| ||||||
| Ashraf et al. [77] | Local Binary Pattern (LBP), SVM | DIARETDB1 | Green channel | 87.48% | 85.99% | 86.15% |