Assistant Diagnosis of Pediatric Pneumonia Based on Vision Transformer

Output size	R50
$112 \times 112$	stdConv， $7 \times 7$ ，64，stride 2
$56 \times 56$	max pool， $3 \times 3$ ，stride 2
$56 \times 56$	$[\begin{array}{l} s t d C o n v, 1 \times 1,64 \\ s t d C o n v, 3 \times 3,64 \\ s t d C o n v, 1 \times 1,256 \end{array}] \times 3$
$28 \times 28$	$[\begin{array}{l} s t d C o n v, 1 \times 1,128 \\ s t d C o n v, 3 \times 3,128 \\ s t d C o n v, 1 \times 1,512 \end{array}] \times 4$
$14 \times 14$	$[\begin{array}{l} s t d C o n v, 1 \times 1,256 \\ s t d C o n v, 3 \times 3,256 \\ s t d C o n v, 1 \times 1,1024 \end{array}] \times 9$
$14 \times 14$	Conv， $1 \times 1$ ，768，stride 1

Table 2. Network structure of R-MSA32 feature extraction stage

View table

Table 2. Network structure of R-MSA32 feature extraction stage

Output size	R-MSA 32
$112 \times 112$	stdConv， $7 \times 7$ ，64，stride 2
$56 \times 56$	max pool， $3 \times 3$ ，stride 2
$56 \times 56$	$[\begin{array}{l} s t d C o n v, 1 \times 1,64 \\ s t d C o n v, 3 \times 3,64 \\ s t d C o n v, 1 \times 1,256 \end{array}] \times 3$
$28 \times 28$	$[\begin{array}{l} s t d C o n v, 1 \times 1,128 \\ s t d C o n v, 3 \times 3,128 \\ s t d C o n v, 1 \times 1,512 \end{array}] \times 4$
$14 \times 14$	$[\begin{array}{l} s t d C o n v, 1 \times 1,256 \\ s t d C o n v, M S A, 256 \\ s t d C o n v, 1 \times 1,1024 \end{array}] \times 3$
$14 \times 14$	Conv， $1 \times 1$ ，768，stride 1

Table 3. Comparison between ViT network model and hybrid ViT network model
View table
Table 3. Comparison between ViT network model and hybrid ViT network model
Model Accuracy Precision Recall
ResNet50 94.70 94.32 96.49
ViT 95.21 96.49 96.94
R50+ViT 96.08 97.42 97.20
ResUNet+R50+ViT 96.58 98.13 97.22
Proposed model 97.26 98.60 97.69

Table 4. Comparison with existing research results
View table
Table 4. Comparison with existing research results
Model Accuracy
$D e n s e N e t^{'}$ 90.46
AlexNet_S 94.87
GIV3 96.77
Proposed model 97.26

Tools

Get Citation

Copy Citation Text

Shuang Zhao, Guohui Wei, Wenhua Zhao, Zhiqing Ma. Assistant Diagnosis of Pediatric Pneumonia Based on Vision Transformer[J]. Laser & Optoelectronics Progress, 2023, 60(4): 0410013

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites