Despeckle filtering in ultrasound video of the common carotid artery

Abstract

Noise reduction is essential for increasing the visual quality or as a preprocessing step for further automated analysis in video sequences and video coding. The objective of this work was to investigate four different video despeckle filtering techniques and evaluate them using visual assessment by two medical experts, texture features analysis, and video quality evaluation metrics. The four proposed video despeckle filtering techniques were evaluated on 10 ultrasound videos of the common carotid artery (CCA). The filters were applied on the whole video frame and in a selected by the user region of interest (ROI) which included the atherosclerotic carotid plaque. The despeckle filters were based on linear filtering (DsFlsmv), hybrid median filtering (DsFhmedian), nonlinear filtering (DsFkuwahara) and speckle reducing anisotropic diffusion (DsFsrad) filtering. Our results showed that, the best video despeckle filtering methods were the linear filter DsFlsmv, followed by the hybrid median filter DsFhmedian. Both filters improved the visual perception evaluation by experts and gave better texture and video quality metrics. Further work on a larger number of videos and by employing additional despeckle filtering techniques is required for the evaluation of video despeckle filtering methods on ultrasound videos of the CCA.

A review of ultrasound common carotid artery image and video segmentation techniques

“…The problems that are associated with the computerassisted border tracing segmentation procedures (also presented in Tables 1, 2, 3, 4) are the following: (1) They do not take into consideration the speckle noise or the image intensity normalization in the ultrasound images [64], or videos [54,55,58]. 2They are sensitive to the initial snake contour [10], or initial seed points, which need to be placed manually [111].…”

“…The two scanners used were the ATL HDI-3000 and the ATL HDI-5000 (Advanced Technology Laboratories, Seattle, USA). Before proceeding with the segmentation, the images and/or videos [54,55,58] were intensitynormalized based on the method introduced in [21]. This improves image compatibility by reducing the variability introduced by different gain settings, different operators, different equipment, and facilitates ultrasound tissue comparability [21,95].…”

Real-time Parallelized Hybrid Median Filter for speckle removal in ultrasound images

“…From [21], [14] and our parametric evaluation, we have noticed wiener filtering preserves edges and other high-frequency information of the image, but requires more computation time than hybrid median filters. Moreover, Frost , Lee, Median and Diffusion despeckle filters perform poorly for ultrasounds since it smoothes edges.…”

“…In [14] Expert 1, a cardiovascular surgeon, and Expert 2, a neurovascular surgeon, preferred hybrid median filter as it preserves edges and important information. Although hybrid median filter provided good results, it isnt applicable in realtime due to its time completion.…”

“…Filters that are used widely in ultrasound imaging include the Frost et al [11], Lee [12], and Kuan et al [13]. Different despeckle filters that were evaluated in [14] [8] showed that hybrid median filter preserved the edges and important information the best with significant speckle removal. Hybrid median filter and other filters performed poorly in real-time ultrasound imaging for its completion time.…”