An Application of Super-Resolution Generative Adversary Networks for Quasi-Static Ultrasound Strain Elastography: A Feasibility Study

Abstract

In this work, a super-resolution approach based on generative adversary network (GAN) was used to interpolate (up-sample) ultrasound radio-frequency (RF) echo data along the lateral (perpendicular to the acoustic beam direction) direction before motion estimation. Our primary objective was to investigate the feasibility of using a GAN-based super-solution approach to improve lateral resolution in the RF data as a means of improving strain image quality in quasi-static ultrasound strain elastography (QUSE). Unlike natural scene photographs, axial (parallel to the acoustic beam direction) resolution is significantly higher than that of lateral resolution in ultrasound RF data. To better handle RF data, we first modified a superresolution generative adversary network (SRGAN) model developed by the computer vision community. We named the modified SRGAN model as super-resolution radio-frequency neural network (SRRFNN) model. Our preliminary experiments showed that, compared with axial strain elastograms obtained using the original ultrasound RF data, axial strain elastograms using ultrasound RF data up-sampled by the proposed SRRFNN model were improved. Based on the Wilcoxon rank-sum tests, such improvements were statistically significant (p < 0.05) for large deformation (3-5%). Also, the proposed SRRFNN model outperformed a commonly-used method (i.e. bi-cubic interpolation used in MATLAB [Mathworks Inc., MA, USA]) in terms of improving axial strain elastograms. We concluded that applying the proposed (SRRFNN) model was feasible and good-quality strain elastography data could be obtained in in vivo tumor-bearing breast ultrasound data. INDEX TERMS Generative adversarial network, motion tracking, super-resolution, quasi-static ultrasound strain elastography.

Unveiling the future of breast cancer assessment: a critical review on generative adversarial networks in elastography ultrasound

“…In a complementary study, He et al. ( 39 ) investigate the suitability of using a GAN-based approach (i.e., SRRFNN) to improve lateral resolution in the radiofrequency (RF) data (i.e., up-sample RF data perpendicular to acoustic beam), consequently improving the elastogram quality in ultrasound strain elastography. However, the V-EUS ( 20 ) generation approach is a preferable end-to-end solution because it generates elastograms directly from conventional B-mode US rather than upsampling the lateral resolution to improve quality.…”

“…Moreover, the authors evaluate the quality of generated elastograms based on improved breast cancer diagnostic accuracy, elastography of deep-seated tumors, and improvement in diagnostic effectiveness of pocket US, which were omitted in the evaluation of AUE-Net for the elastography of thyroid nodule. In a complementary study, He et al (39) investigate the suitability of using a GAN-based approach (i.e., SRRFNN) to improve lateral resolution in the radiofrequency (RF) data (i.e., up-sample RF data perpendicular to acoustic beam), consequently improving the elastogram quality in ultrasound strain elastography. However, the V-EUS (20) generation approach is a preferable end-to-end solution because it generates elastograms directly from conventional B-mode US rather than upsampling the lateral resolution to improve quality.…”

Deep learning in ultrasound elastography imaging: A review

“…123,180 GANs have been used for super resolution in ultrasound elastography. 181 Deep learning methods also have the ability to create synthetic elastography images from ultrasound B-mode images. 83 Image fusion is another feature of deep learning, which is a future scope in ultrasound elastography by now, but is already popular for other imaging modalities.…”

“…Many super resolution models such as FUS‐Net are based on extensive down‐ and up‐sampling of the images 123,180 . GANs have been used for super resolution in ultrasound elastography 181 . Deep learning methods also have the ability to create synthetic elastography images from ultrasound B‐mode images 83 .…”