Figure 1.
The basic architecture of a convolutional neural network
Figure 2.
Convolution Process Diagram
Figure 3.
Pooling Process Diagram
Figure 4.
VGG16 Network Architecture Diagram
Figure 5.
Residual Connection Structure Diagram
Figure 6.
Transformer Architecture Diagram
Figure 7.
Self-Attention Mechanism Computation Diagram
Figure 8.
ViT Architecture Diagram
Figure 9.
DeiT Architecture Diagram.
TABLE I.
EXPERIMENTAL DATASET TABLE
| Training Set | Validation Set | Test Set | |
|---|---|---|---|
| lung_aca | 3500 | 750 | 750 |
| lung_scc | 3500 | 750 | 750 |
| lung_n | 3500 | 750 | 750 |
Figure 10.
Randomly selected training sample images
Figure 11.
Accuracy and Loss Curves of theVgg16
Figure 12.
Accuracy and Loss Curves of the ResNet50
Figure 13.
Accuracy and Loss Curves of the DeiT
Figure 14.
Confusion Matrix
TABLE II.
Prediction Metrics for Different Models
| Model | Acc(%) | Average Precision(%) | Average Recall(%) | Average F1-Score(%) |
|---|---|---|---|---|
| Vgg16 | 98.49 | 98.49 | 98.49 | 98.49 |
| Resnet50 | 97.51 | 97.51 | 97.51 | 97.51 |
| DeiT | 99.96 | 99.96 | 99.96 | 99.96 |
Figure 15.
Random Test Plot