Table 3.
Comparative study with CNN-based approaches
| Method/year | Preprocessing | Auto/semi-auto | Dimension | Method | Accuracy |
|---|---|---|---|---|---|
|
[50] (Ge et al. 2018) |
- | Auto | 2D | CNN | 90.4% |
|
[22] (Iram et al.2018) |
- | Semi-auto | 3D | AlexNet-LSTM | 71% |
|
[22] (Iram et al.2018) |
- | Semi-auto | 3D | ResNet-LSTM | 71% |
|
[22] (Iram et al.2018) |
- | Semi-auto | 3D | VGGNet-LSTM | 84% |
|
[51] (Pan,et al.,2015) |
Intensity normalization | Auto | 2D | CNN | 60% |
| Proposed approach | Intensity normalization/contrast enhancement | Auto | 3D | Deep CNN | 96.49% |
Entries in bold indicate the proposed method