UW-Madison ECE 533 - Despeckle Filtering in Medical Ultrasound Imaging

Unformatted text preview:

ECE 533 “Digital Imaging Processing” Course Project Report: Despeckle Filtering in Medical Ultrasound Imaging Hairong Shi (1) Xingxing Wu (2) (1) Department of Medical Physics, University of Wisconsin-Madison (2) Department of Electrical and Computer Engineering, University of Wisconsin-Madison Dec. 12, 2003Percentage of each person’s work: Hairong Shi: 50% (Wiener Filter Method, Anisotropic Diffusion Method, Simulation of inclusion phantoms) Xingxing Wu: 50% (Wavelet Filter Method, K-distribution based Adaptive Filtering Method)I Introduction The medical Ultrasound B-scan (brightness scan) echo imaging is acquired by summation of the echo signals from ultrasound scatterers in the ultrasound beam range. The scatterers are from structures, tissue interfaces and tissue microstructures etc. in the body, these scatterers are locally correlated. And the coherent summation of signals include the structures and interfaces information which are useful for diagnosis purpose, as well as some locally correlated multiplicative noises from scatterers smaller than ultrasound beam wavelength (resolution size), which corrupts medical ultrasound imaging and makes visual observation difficult. These noises are commonly called “speckles”. Even though in some cases the speckle are essential information to track features, many cases the speckle noise deteriorates the image quality, degrades the fine details and edge definition. It also limits the contrast resolution, limiting the detectability of small, low contrast lesions in body. Speckle is always considered as a primary source of medical ultrasound imaging noise, and it should be filtered out. A simple example is shown in Fig. 1 to demonstrate the impact of speckle noise on information content (image from Duke University). The object of interest is a hypoechoic lesion of 5 mm diameter with -9 dB contrast. The echogenicity map corresponding to this object is shown in the top left panel of Fig.1. The scattering function represents the population of sub-resolution scatterers being imaged, which is shown in the top right panel of Fig. 1. The RF echo data is shown in the lower left panel. It is zero-mean and thus does not show a map of local echo magnitude. The low right panel shows the envelope detected image, which produces the desired image of echo magnitude. From the images, it is easy to see how speckle noise obscures the information in the image. Usually in clinical application, the B-mode images shown on ultrasound machine monitor are logarithm compressed envelope-detected image, as shown in the bottom right image shown in Fig. 1. While for research application, usually Radio-frequency (RF) data are collected directly from ultrasound machine, which can be considered as “raw” data from transducer and amplifier circuits. To convert RF data into B-mode images, the data are first Hilbert transformed to detect envelope, and then do logarithm. Due to such operation, the multiplicative speckle is converted into additive noise after logarithm compression, and the signal created has a Rayleigh amplitude PDF: ()0,2exp222>−= aaaapAσσ For images with speckles, removing speckles while not affecting important features is the purpose of despeckle. Many methods are proposed to alleviate the speckles. Spatial compounding method attempts to reduce the noise by averaging several images (Li and O’Donnell 1994). Filtering methods are practical alternatives. In this project, we implemented and tested several despeckle filters.Figure 1. The ultrasound image of a hypoechoic lesion of 5 mm diameter with -9 dB contrast. (Top Left) The echogenicity map. (Top Right) The scattering function. (Lower Left) RF echo data. (Lower Right) Envelope detection Image. II Methods and Results (0) Test images In this project, we use the following figures to test our filters. (A) 4 simulated inclusion phantoms with different contrast. The simulation program is based on a previous research in our laboratory (Li and Zagzebski, 1999). The simulated ultrasound beam has center frequency 3MHz, band width 40%, no attenuation. We simulate phantom with three different size inclusions embedded in the background gelatin, the signal contrast (inclusion to background) for four phantoms are 10dB, 5dB, -5dB and -10dB respectively. The images of these 4 phantoms are shown below.cmcm10dB0 2 4 601234567cmcm5dB0 2 4 601234567(a) (b) cmcm-5dB0 2 4 601234567cmcm-10dB0 2 4 601234567(c) (d) Figure 2 Four simulated phantoms with different contrast (a) 10dB, (b) 5dB, (c) -5dB, (d) -10dB (B) We acquired RF data of a uterine phantom from Aloka SSD2000 Medical Ultrasound System. The transducer is a linear transducer with center frequency 7.5MHz, sampling rate is 100MHz, the image then is down sampled to reduce the computation time. The phantom shown has a uterine tube and three lesions at the wall.cmcmUterine Phantom0 2 4012345678 Figure 3 B-mode image of a Uterine phantom (C) An in-vitro B-mode image for a plaque in human carotid artery. The plaque is removed by surgery operation, and embedded into gelatin phantom. The RF data is also acquired with Aloka SSD 2000 system, linear transducer with center frequency 7.5MHz, sampling rate 100MHz. The B-mode image is shown in Fig. 4. The plaque is at the center of the image. cmcmCarotid Artery Plaque0 0.5 1 1.5 2 2.5 300.511.522.53 Figure 4 B-mode image of a carotid artery plaqueFrom Figures above we can see the image qualities for B-mode image are usually very poor. And thus the despeckle filtering is necessary. In the following section we will show the filtering results by 4 methods: (1) Wiener filter; (2) Anisotropic diffusion filter; (3) K-distribution based adaptive filter; (4) Wavelet filter. (1) Wiener Filter Wiener filter equation in frequency domain is written as: wwssssSSggSgW+⋅=** Where g is the filter convolve the input image, Sww is the power spectrum of the noise, Sss is the power spectrum of the input image. In this problem, we only assume the input image is only added with noise, so the filter g=1 in frequency domain. The Wiener filter now is simplified into wwssssSSSW+= The power spectrum of the input image Sss is unknown. An easy way is to model the power spectrum of the input image as ()2222yxsssSµµσ+= Where σs2 is the mean variance of the input image S, which is also unknown. However, we usually use the mean variance of the


View Full Document

UW-Madison ECE 533 - Despeckle Filtering in Medical Ultrasound Imaging

Documents in this Course
Load more
Download Despeckle Filtering in Medical Ultrasound Imaging
Our administrator received your request to download this document. We will send you the file to your email shortly.
Loading Unlocking...
Login

Join to view Despeckle Filtering in Medical Ultrasound Imaging and access 3M+ class-specific study document.

or
We will never post anything without your permission.
Don't have an account?
Sign Up

Join to view Despeckle Filtering in Medical Ultrasound Imaging 2 2 and access 3M+ class-specific study document.

or

By creating an account you agree to our Privacy Policy and Terms Of Use

Already a member?