Segmentation with Speckle Reduction and Superresolution by Deep Leaning for Human Ultrasonic Echo Image

Ultrasound (US) image diagnosis is widely used for detection and treatment of human malignant tissues. Physicians perform differentiation of tissues through interpreting ultrasound echo images morphologically. However, the ultrasound image always comes with speckles, which makes segmentation of a target tissue difficult. Recently, a deep learning (DL) approach becomes a new way for picture denoising instead of signal processing. In this report, we use the DL denoising to reduce the US speckles. Subsequently, we perform DL segmentation well known for other medical images. In order to further increase the segmentation accuracy, we also perform DL superresolution. The DL superresolution is also well known for a picture and however, not so for an echo image. The target segmentation tissue is a carotid artery, specifically a lumen. To verify the feasibilities of our approaches, simulations and in vivo experiments are performed.Clinical Relevance- Method effectiveness is confirmed for in vivo data.