Table 1.
Summary of previous works in keratoconus classification since 2012.
| Authors | Year | Method | Dataset | Inputs | Accuracy | Feature selection |
|---|---|---|---|---|---|---|
| Al-Timemy et al. [8] | 2021 | SqN, AlN, SfN, MbN | 2136 images | N.A | 92.2% to 94.8% | N.A |
| Kamiya et al. [9] | 2021 | CNN | 3390 images | N.A | 78.5% | N.A |
| Jiménez-García et al. [10] | 2021 | TDNN | 743 images | 6 | N.A | Yes |
| Kuo et al. [11] | 2020 | VGG16, InceptionV3, ResNet152 | 354 images | N.A | 93.1%, 93.1%, 95.8% | N.A |
| Cao et al. [22] | 2020 | RF, SVM, KNN, LR, LDA, LaR, DT, MPAN | 88 eyes | 11 | 87%, 86%, 73%, 81%, 81%, 84%, 80%, 52% | Yes |
| Lavric et al. [24] | 2020 | 25 classifiers | 3151 images | 8 | 62% to 94% | SS, FRank |
| Velázquez-Blázquez et al. [23] | 2020 | LR | 178 eyes | 5 | 73% | X 2, Kruskal-Wallis |
| Lavric and Valentin [12] | 2019 | CNN | 3000 | 180 × 240 × 3 (images) | 99.33% | Yes |
| Issarti et al. [13] | 2019 | FNN | 851 | 141 × 141 (images) | 96.56% | NCAFS |
| Salem and Solodovnikov [14] | 2019 | RF | 500 | N.A | 76% | Yes |
| Hallett et al. [15] | 2019 | BNN | 124 | 29 | 73% (supervised) 80% (unsupervised) | PCA |
| Luna et al. [16] | 2019 | BNN | 60 | 16 | 100% | N.A |
| Kamiya et al. [17] | 2019 | CNN | 543 | 6 × 224 × 224 (image) | 99.1% | N.A |
| Yousefi et al. [5] | 2018 | UnML | 3156 | 420 | N.A | PCA NonLinear_tSNE |
| Hidalgo et al. [18] | 2017 | SVM | 131 | 25 | 92.6% to 98% | N.A |
| Ali et al. [21] | 2017 | SVM | 40 | 12 | 90% | N.A |
| Smadja et al. [19] | 2013 | DT | 372 | 55 | N.A | N.A |
| Arbelaez et al. [20] | 2012 | SVM | 3502 | 7 | 98.2% | N.A |