Sparse-View Photoacoustic Image Quality Enhancement Based on a Modified U-Net

Tong Wang; Wende Dong; Kang Shen; Songde Liu; Wen Liu; Chao Tian

doi:10.3788/LOP202259.0617022

Laser & Optoelectronics Progress, Volume. 59, Issue 6, 0617022(2022)

Sparse-View Photoacoustic Image Quality Enhancement Based on a Modified U-Net

Tong Wang¹, Wende Dong², Kang Shen^3,4, Songde Liu^3,4, Wen Liu¹, and Chao Tian^3,4、*

Author Affiliations

¹School of Physical Science, University of Science and Technology of China, Hefei , Anhui 230026, China

²College of Automation Engineering, Nanjing University of Aeronautics and Astronautics, Nanjing , Jiangsu 211106, China

³School of Engineering Science, University of Science and Technology of China, Hefei , Anhui 230026, China

⁴Key Laboratory of Precision Scientific Instrumentation of Anhui Higher Education Institutes, Hefei , Anhui 230026, China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(9)

Fig. 1. Architecture of the modified U-Net. (a) Conventional U-Net network; (b) modified skip connection network

Download full size

Fig. 2. Reconstruction results on simulation dataset. (a) (e) Images reconstructed by an iterative algorithm; (b) (f) U-Net outputs; (c) (g) modified U-Net outputs; (d) (h) reference images

Download full size

Fig. 3. Comparison of profiles labeled by arrows in Fig. 2 (d) and (h). (a) N=32; (b) N=128

Download full size

Fig. 4. Reconstruction results of a cross section of the liver of a mouse on experimental dataset. (a) Position of imaging; (b) (f) images reconstructed by an iterative algorithm; (c) (g) U-Net outputs; (d) (h) modified U-Net outputs; (e) demonstration of liver area; (i) reference image

Download full size

Fig. 5. Reconstruction results of a cross section of the spleen of a mouse on experimental dataset. (a) Position of imaging; (b)(f) images reconstructed by an iterative algorithm; (c) (g) U-Net outputs; (d) (h) modified U-Net outputs; (e) demonstration of spleen area; (i) reference image

Download full size

Fig. 6. Comparison of training losses and validating losses during training process. (a) Simulation dataset, element number N=32; (b) simulation dataset, element number N=128; (c) experimental dataset, element number N=32;(d) experimental dataset, element number N=128

Download full size

Table 1. Quantitative comparisons of different methods on simulation dataset
View table
Table 1. Quantitative comparisons of different methods on simulation dataset
Item Iteration U-Net Modified U-Net
N=32 N=128 N=32 N=128 N=32 N=128
PSNR 22.18 40.53 31.23 48.37 34.23 50.90
SSIM 0.231 0.884 0.975 0.996 0.987 0.996

Table 2. Quantitative comparisons of different methods on experimental dataset
View table
Table 2. Quantitative comparisons of different methods on experimental dataset
Item Iteration U-Net Modified U-Net
N=32 N=128 N=32 N=128 N=32 N=128
PSNR 20.03 26.47 30.08 34.74 31.17 35.16
SSIM 0.117 0.383 0.707 0.827 0.710 0.836

Table 3. Results of ablation experiment
View table
Table 3. Results of ablation experiment
Item U-Net+L_mix Modified U-Net+L_MSE Modified U-Net+L_mix
PSNR 34.74 35.13 35.16
SSIM 0.827 0.834 0.836

Tools

Get Citation

Copy Citation Text

Tong Wang, Wende Dong, Kang Shen, Songde Liu, Wen Liu, Chao Tian. Sparse-View Photoacoustic Image Quality Enhancement Based on a Modified U-Net[J]. Laser & Optoelectronics Progress, 2022, 59(6): 0617022

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites