File Exchange

image thumbnail

Efficient B-mode Ultrasound Image Reconstruction Using CNN

version 1.0.1 (54.3 MB) by Shujaat Khan
Efficient B-mode Ultrasound Image Reconstruction from Sub-sampled RF Data using Deep Learning


Updated 26 Nov 2018

GitHub view license on GitHub

Yoon, Yeo Hun, Shujaat Khan, Jaeyoung Huh, and Jong Chul Ye. "Efficient B-mode Ultrasound Image Reconstruction from Sub-sampled RF Data using Deep Learning." IEEE transactions on medical imaging (2018).
MatConvNet (matconvnet-1.0-beta24)
Please run the matconvnet-1.0-beta24/matlab/vl_compilenn.m file to compile matconvnet.
There is instruction on ""
Please run the installation setup (install.m) and run some training examples.
Trained network
Trained network for 'SC2xRX4 (down-sampling) CNN' is uploaded.
Test data
Test data file is placed in 'data\cnn_sparse_view_init_multi_normal_dsr2_input64' folder.
The dimension of data are as follows -- Test_data = 64x384x1x2304 (channel x scanline x frame x depth)
To perform a test using proposed algorithm

-> Use 'DNN4x1_TestVal' as input data


-> You will get the reconstructed RF data in the 'data\cnn_sparse_view_init_multi_normal_dsr2_input64' directory.

-> Using standard delay-and-sum (DAS) beam-forming code construct a B-mode image. For our experiments we used a DAS beam-forming code provided by (Alpinion Co., Korea). A similar code can be downloaded from ('').

Cite As

Yoon, Yeo Hun, et al. “Efficient B-Mode Ultrasound Image Reconstruction from Sub-Sampled RF Data Using Deep Learning.” IEEE Transactions on Medical Imaging, Institute of Electrical and Electronics Engineers (IEEE), 2018, pp. 1–1, doi:10.1109/tmi.2018.2864821.

View more styles

Comments and Ratings (0)



- citation information update

MATLAB Release Compatibility
Created with R2018b
Compatible with any release
Platform Compatibility
Windows macOS Linux