2012 Summer Project Week Breakout Session: Ultrasound: Denoising
US Speckle Parameter Estimate
Module Type & Category
Authors, Collaborators & Contact
- Author: Kunlin Cao and Kedar Patwardhan (GE Research)
- Contact: email@example.com
The USSpeckleParameterEstimate filter performs an information-theoretic based probabilistic modeling of speckles in Ultrasound images. Intensities of speckles in Ultrasound images are modeled using Rayleigh distribution or normalized Fisher-Tippett distribution.
Speckles are formed from coherent accumulation of random scattering in a cell of ultrasound beam. The pixel intensity of the speckle in the inphase/quadrature image follows a zero-mean Gaussian distribution. Then envelope detection of the inphase/quadrature image is performed to produce a real image/magnitude image, whose intensity corresponds to a Rayleigh distribution. Due to the large dynamic range of the magnitude image, log-compress is usually performed to produce an image suitable for display. A pixel’s intensity in the resulting image follows the normalized Fisher-Tippett distribution. The maximum likelihood estimate (MLE) of the distribution parameters in a neighborhood are calculated to describe the image information and with less speckle effects.
A complete description of the algorithm may be found in:
G. Slabaugh, G. Unal, M. Wels, T. Fang, and B. Rao. Statistical Region-based Segmentation of Ultrasound Images. Ultrasound in Medicine and Biology, 35(5):781 – 795, 2009.
Examples, Use Cases & Tutorials
- This module is used to reduce speckles in an ultrasound image. Usually the normalized Fisher-Tippett distribtuion is used to model the statistical parameter.
- A command line example of running this module is
USSpeckleParameterEstimate --neighborhood 1,1,1 inputUltrasoundImage.nhdr outputSmoothedImage.nhdr
The screenshots on the top illustrate how to use the filter and the expected results in two cases. The inputs are ultrasound images, and outputs are MLE estimate of normalized Fisher-Tippett distribution parameters (smoothed image).
Quick Tour of Features and Use
- Input/output panel:
User selects an ultrasound image as input upon which the estimate filter will apply. User creates an output image to store the result. Upon completion, the output image will be displayed in the viewing panel.
- Parameters panel:
User specifies the neighborhood radius in three dimensions (comma separated without spaces). User chooses the distribution type for modeling (Rayleigh or Fisher-Tippett).
- Viewing panel:
Upon completion, the estimate filter result is displayed in the viewing panel.
None other than the core Slicer3 modules for image IO and display.
- G. Slabaugh, G. Unal, M. Wels, T. Fang, and B. Rao. Statistical Region-based Segmentation of Ultrasound Images. Ultrasound in Medicine and Biology, 35(5):781 – 795, 2009.
- Z. Wang, G. Slabaugh, G. Unal, and T. Fang. Information-Theoretic Feature Detection and Its Application to Registration of Ultrasound Images. International Journal of Intelligent Control and Systems, 13(2):136–145, 2008.