Table 3.
Comparison among different methods based on the average sensitivity, specificity, accuracy, precision, and F1 Score.
| Reference | Method | Sensitivity | Specificity | Accuracy | Precision | F1 Score |
|---|---|---|---|---|---|---|
| [24] | Convolutional neural network and CLAHE framework | - | - | 98.96% | - | - |
| [25] | Convolution neural networks | - | - | 96.00% | - | - |
| [26] | Machine learning approach and deep-learning-based | 96.37% | - | 96.33% | 96.39 | 96.38% |
| [27] | Deep Learning Method | 97.00% | 97.00% | 97% | 97.00% | |
| [28] | Deep Learning with Bayesian–Gaussian-Inspired Convolutional Neural Architectural Search | 93.00% | - | 97.92% | 97% | 97.00% |
| [29] | Hybrid principal component analysis network and extreme learning machine | 99.12% | 99.38% | 98.97% | 98.87% | 98.84% |
| [30] | Convolutional Neural Network | 99.00% | - | 99.00% | 98.6% | 98.8% |
| [31] | Transfer learning with class-selective image processing | - | - | 98.40% | ||
| [32] | Partial self-supervised learning | 95.74% | 80.95% | 93.04% | 95.74% | 95.74% |
| Proposed Method | CNN based on SqueezeNet and GOA | 99.34% | 99.41% | 99.12% | 98.91% | 98.94% |