Table 5.
Comparison of classification accuracy among five different DCNN models
| Index | AlexNet | ZFNet | VGG16 | GoogLeNet | Proposed method |
|---|---|---|---|---|---|
| Classification accuracy | 78.6% (44/56) | 76.8% (43/56) | 76.8% (43/56) | 75.0% (42/56) | 92.9% (52/56) |
| Sensitivity | 76.7% (23/30) | 76.7% (23/30) | 83.3% (25/30) | 73.3% (22/30) | 93.3% (28/30) |
| Specificity | 80.8% (21/26) | 76.9% (20/26) | 69.2% (18/26) | 76.9% (20/26) | 92.3% (24/26) |
| PPV | 82.1% (23/28) | 79.3% (23/29) | 75.8% (25/33) | 78.6% (22/28) | 93.3% (28/30) |
| NPV | 75.0% (21/28) | 74.1% (20/27) | 78.3% (18/23) | 71.4% (20/28) | 92.3% (24/26) |