Despeckling of Medical Ultrasound Images: A Technical Review

Abstract

Abstract-Acquisition of digital image and preprocessing methods plays a vital role in clinical diagnosis. The ultrasound medical images are more popular than other imaging modalities, due to portable, adequate, harmless and cheaper nature of it. Because of intrinsic nature of speckle noise (signal based noise), ultrasound medical image leads to degradation of the resolution and contrast of the image. Reduction of this signal based noise is helpful for the purpose of visualization of the ultrasound images. The low quality of image is considered as a barrier for the better extraction of features, recognition, analysis and detection of edges. Because of which inappropriate diagnosis may be done by doctor. Thus, speckle noise reduction is essential and preprocessing step of ultrasound images. Analysts survey manifold reduction methods of speckle noise, yet there is no exact method that takes all the limitations into account. In this review paper, we compare filters that are Lee, Frost, Median, SRAD, PMAD, SRBF, Bilateral, Adaptive Bilateral and Multiresolution on medical ultrasound images. The results are compared with parameter PSNR along with the visual inspection. The conclusion is illustrated by filtered images and data tables.

RobUSt–An Autonomous Robotic Ultrasound System for Medical Imaging

“…While conventional US image analysis involves a clinician and requires experience, automatic segmentation methods can improve accuracy and clinical outcome [22] or enable real-time systems to control medical instruments like needles [23]- [25], catheters [26], or drills [27]. However, US images inherently contain many imaging artifacts and have low signal-to-noise ratio so filtering of noise from the signal is one of the main perceived challenges [28].…”

Advanced robotics for magnetic manipulation : the mobile magnetic vascular surgeon

“…Analysis of robotic US images opens up new diagnostic possibilities, as information about the pose of the probe corresponding to each analysed image allows for the creation of a 3D data set of anatomical landmarks. However, US images inherently contain many imaging artifacts and have low signal-to-noise ratio [326]. Thus, one of the challenges associated with the analysis of US images is the filtering of noise from the signal, which preserves fine details that may have characteristics similar to noise [327].…”