Low-Dose CT Image Denoising Based on Error Modulation Module

Fig. 5. Comparison of denoising effects on AAPM-1 mm-L286-30 slices. (a) (g) (m) NDCT; (b) (h) (n) (s) LDCT; (c) (i) (o) (t) BM3D; (d) (j) (p) (u) RED-CNN; (e) (k) (q) (v) DU-GAN; (f) (l) (r) (w) our method

Download full size

Fig. 6. Comparison of denoising effects on AAPM-1 mm-L286-440 slices. (a) (g) (m) NDCT; (b) (h) (n) LDCT; (c) (i) (o) BM3D; (d) (j) (p) RED-CNN; (e) (k) (q) DU-GAN; (f) (l) (r) our method

Download full size

Fig. 7. Comparison of HU value difference. (a) NDCT; (b) LDCT; (c) archetype; (d) add error modulation module

Download full size

Fig. 8. Comparison of noise reduction results of single loss and compound loss. (a) (e) NDCT; (b) (f) LDCT; (c) (g) MSE; (d) (h) compound loss

Download full size

Table 1. Mean value comparison of the first 275 groups of pictures
View table
Table 1. Mean value comparison of the first 275 groups of pictures
Denoising method PSNR /dB SSIM
LDCT 24.47±1.76 0.7763±0.0605
BM3D 27.96±0.98 0.8976±0.0192
RED-CNN 28.82±1.53 0.9599±0.0106
DU-GAN 31.50±1.22 0.9591±0.0121
Our method 32.86±1.47 0.9756±0.0098

Table 2. Mean value comparison of the last 250 groups of pictures
View table
Table 2. Mean value comparison of the last 250 groups of pictures
Denoising method PSNR /dB SSIM
LDCT 25.49±2.15 0.4310±0.0908
BM3D 27.67±0.95 0.8763±0.0084
RED-CNN 27.92±1.53 0.9363±0.0136
DU-GAN 28.39±1.13 0.9202±0.0219
Our method 29.18±1.01 0.9603±0.0062

Table 3. Comparison results of ablation experiments
View table
Table 3. Comparison results of ablation experiments
Archetype EMM PSNR SSIM
√ – 31.28 0.9683
√ √ 32.86 0.9756

Tools

Get Citation

Copy Citation Text

Xiaohe Zhao, Ping Chen, Jinxiao Pan. Low-Dose CT Image Denoising Based on Error Modulation Module[J]. Laser & Optoelectronics Progress, 2025, 62(8): 0817001

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites