An Edge-Preserving Markov-Random-Fields Model for Speckle Removal in Ultrasound Images
S Tsantis1, P Katsakiori2, D Karnabatidis2, A Skouroliakou3, D Mihailidis4, GC Kagadis1*, (1) University of Patras, Rion, Ahaia, GR (2) University of Patras, Rion, Ahaia, GR (3) Technical Educational Institute of Athens, Aigaleo, Attiki, GR (4) Charleston Radiation Therapy Cons, Charleston, WV.SU-E-U-7 Sunday 3:00PM - 6:00PM Room: Exhibit Hall
Purpose: Contemporary ultrasound (US) systems have gained the confidence of medical community among other imaging modalities such as CT and MRI. However, US images carry a granular pattern, so called speckle, which constitutes a major image quality degradation factor. A new Markov random fields (MRF) model is proposed for the detection and removal of speckle noise in ultrasound images.
Methods: 20 ultrasound images were analyzed. The MRF model design comprised two distinct sources of information: (a) the likelihood function that characterizes the contrast likelihood at a site, and (b) the a priori knowledge derived from the wavelet transform of the US image. The likelihood probability density function (pdf) is approximated by the combination of the intensity distribution and the wavelet transform modules (WTM) values of individual regions in the image. The a priori knowledge or contextual information is described by the positions and the angle vector of the WTM values. The combination of these sources builds the MRF model and provides an accurate edge map that is employed in the speckle reduction procedure that follows.
Results: The proposed MRF model addresses the speckle problem that dominates US imaging. Speckle noise is reduced significantly while all edges remained intact. It exhibited similar results in terms of speckle index (SI), signal-to-mean-square-error ratio (S/mse) and edge preservation index β compared with the commercially available denoising packet introduced by General Electric termed as Speckle Reduction Imaging (SRI).
Conclusion: Experimental results have demonstrated that an efficient speckle suppression algorithm can improve the overall image quality, which in turns could improve the decision-making procedure in ultrasound imaging.