Improving Ultrasound Imaging Using Deep Adaptive Network Loss (DANL) Techniques

Abstract

Higher image quality and reduction of imaging tools in recent days are two latest advances in ultrasonic imaging. Poor quality imaging leads to more complex in ultrasound imaging, only expensive equipment gives high quality imaging. The images are created by the device which is compact with lower spatial resolution, low contrast, and larger noise. These are problems in ultrasound imaging. This paper addressing the issues and problems in ultrasound imaging. Generative adversarial network (GAN) is proposed and combining the quality loss function.  The quality loss function has two stages are implemented in this work.(1) Classical quality-aware loss; (2) adaptive deep network loss is used to determine the quality score and high content quality arrangement among the input image and renovated image. The proposed algorithm improves the poor quality image, efficiency and flexibility in diagnostics, spatial resolution, and contrast of the image. Speckle noise is removed and the execution time is low comparing to the existing methods.