search
Contact Us
HOME

  • Crossref logo
  • Crossref Similarity Check logo
Journal Search Engine

to

Search Advanced Search Adode Reader(link)
Download PDF Export Citaion korean bibliography PMC previewer
ISSN : 2288-3509(Print)
ISSN : 2384-1168(Online)
Journal of Radiological Science and Technology Vol.46 No.6 pp.475-484
DOI : https://doi.org/10.17946/JRST.2023.46.6.475

Comparative Evaluation of Filters for Speckle Noise Reduction in a Clinical Liver Ultrasound Image

Hajin Kim1), Youngjin Lee2)
1)Department of Health Science, General School of Gachon University
2)Department of Radiological Science, Gachon University

This paper is supported by the academic activities grant from the society of Incheon Radiological Technologists Association (IRTA) of the
Korean Radiological Technologists Association (KRTA) in 2023.


Corresponding author: Youngjin Lee, Department of Radiological Science, Gachon University, 191 Hambakmoe-ro, Yeonsu-gu, Incheon 21936, Republic of Korea / Tel: +82-32-820-4362 / E-mail: yj20@gachon.ac.kr
10/10/2023 08/11/2023 15/11/2023

Abstract


This study aimed to compare filters for reducing speckle noise in ultrasound images using clinical liver images. We acquired the clinical liver ultrasound images, and noisy images were obtained by adding 0.01, 0.05, 0.10, and 0.50 intensity levels of speckle noise to the liver images. The Wiener filter, median modified Wiener filter, gamma filter, and Lee filter were designed for the noisy images by setting window sizes at 3×3, 5×5, and 7×7. The coefficient of variation (COV) and contrast to noise ratio (CNR) were calculated to evaluate noise reduction and various filters. Moreover, the filter with the highest image quality was selected and quantitatively compared to a noisy image. As a result, COV and CNR showed the noise improved result when the Lee filter was applied. Furthermore, the Lee filter image with a window size of 7×7 was noted to possess approximately a minimum of 1.28 to a maximum of 3.38 times better COV and a minimum of 2.18 to a maximum of 5.50 times better CNR than the noisy image. In conclusion, we confirmed that the Lee filter was effective in reducing speckle noise and proved that an appropriate window size needs to be set considering blurring.


Liver Ultrasound Image , Liver Blood Vessel , Speckle Noise Reduction , Lee Filter , Window Size , Quantitative Evaluation

간 초음파 영상에서의 스페클 노이즈 제거를 위한 필터들의 비교 평가

김하진1), 이영진2)
1)가천대학교 일반대학원 보건과학과
2)가천대학교 방사선학과

초록

간 초음파 영상 , 간 혈관 , 스페클 노이즈 제거 , Lee 필터 , 윈도우 크기 , 정량적 평가

    Ⅰ. Introduction

    According to a report by the World Health Organization (WHO), liver disease is one of the causes of approximately 1.1 million death worldwide[1]. According to the 2019 Death Statistics released by the Ministry of Health and Welfare, liver disease has consistently ranked among the top 10 leading causes of death for the past decade, from 2009 to 2019. Particularly, liver cancer is the second most common cancer in Korea, accounting for 20.6% of death, owing to its high frequency and difficult treatment[2]. Currently, medical devices such as computed tomography (CT), magnetic resonance imaging (MRI), and ultrasound, are used to evaluate and diagnose liver diseases in the clinical field[3-5]. Liver ultrasound is essential in determining the shape and size of blood vessels and tissues inside and outside the liver[6-8].

    Liver ultrasound has the advantage of evaluating the liver function in the form of real-time images without exposure to the radiation or contrast agents, and it is non-invasive[9,10]. In addition, liver ultrasound serves a very important role in the early detection of liver disease, such as liver cancer, cirrhosis, and liver cysts. Hence, it is currently the most actively used method for the prevention and diagnosis of most liver diseases[11-13]. Because liver function is determined based on the inflow and regulation of blood supplied by blood vessels, when examining the liver via ultrasound, the ability to differentiate between tissues and blood vessels is critical to assess the blood supply to the liver[14,15]. In particular, certain liver diseases cause abnormalities in the liver veins, such as blood flow status, vessel size, and blood clots, and hepatic vein examination is useful for the early detection and diagnosis of the early stages of liver diseases, such as liver cancer; therefore, observing the hepatic veins on liver ultrasound provides important information[16,17].

    However, ultrasound imaging diagnostics suffer from lower image quality than other medical diagnostic modalities, such as CT, MRI, and X-ray imaging, and the main cause of this problem is small speckle noise. Speckle noise occurs in the form of multiplicative noise when ultrasound pulses randomly interfere with small particles or objects at a scale comparable to that of sound waves because of diffuse scattering[18,19]. In ultrasound examination, noisy images cause difficulty to distinguish between blood vessels and tissue structures and to detect abnormalities in the liver tissue because the noise inside the blood vessels interferes with the signal, thereby distorting the results of the examination. Thus, when examining the liver, a technique to reduce to the noise inside the blood vessels plays a critical role in improving the accuracy of the examination results[20].

    Therefore, speckle noise reduction is one of the important preprocessing steps required for ultrasound image analysis and processing. Various techniques have been proposed for speckle noise reduction. Studies conducted earlier involved reducing speckle noise by improving the hardware of ultrasound transducer, thereby significantly impacting the quality of the ultrasound images; however, they had technical limitations, such as cost and complexity[21,22]. By contrast, software techniques, that is, filtering techniques are the most widely used, because they can be applied to images acquired using existing ultrasound equipment without modifying the existing hardware, are simple to implement, and have significant economic benefits[23,24].

    Therefore, in this study, we applied widely used filters, including Wiener filter, MMWF, gamma filter, and Lee filter to reduce speckle noise in liver ultrasound images and quantitatively evaluated the degree of improvement in image quality to identify the most effective filter. In addition, to evaluate the efficiency of the filters according to the noise intensity, we added various noise intensities and changed the smoothing parameter, window size of the filters to quantitatively compare the performance of the filters in reducing speckle noise.

    Ⅱ. Materials and methods

    1. Acquisition of noisy image

    This study was conducted under the ethical guidelines of the Helsinki Declaration and received prior approval for exemption from review by the Gachon University Institutional Review Board (IRB No.: 1044396-202204- HR-075-01). We acquired a clinical liver ultrasound image to compare the noise reduction performance of the filters in reducing speckle noise in liver ultrasound images. The clinical liver ultrasound image was obtained using a convex probe with a 3.5 MHz frequency through a subcostal scan. Subsequently, to quantitatively evaluate the efficacy of the filters based on varying levels of noise intensity, we introduced speckle noise with intensities of 0.01, 0.05, 0.10, and 0.50 to the acquired liver ultrasound images using the MATLAB software (ver. 2023a, MATHWORKS, Boston, United States of America).

    2. Speckle noise reduction filter modeling

    To reduce speckle noise, a Wiener filter, MMWF, gamma filter, and Lee filter were applied to noisy images of each intensity from the clinical liver ultrasound images. The smoothing parameters of each filter, that is window sizes, were set to 3×3, 5×5, and 7×7. After applying these to the noisy images, a comparative evaluation of the filters was conducted based on a quantitative evaluation.

    The gamma filter is a maximum a posteriori (MAP) filter based on the Bayesian analysis of the image statistics. A gamma filter is used to minimize the loss of texture information and uses the coefficient of variation (COV) and contrast to noise ratio (CNR), the probability density function of which determines the smoothing process[25]. The formula for the gamma filter is as follows:

    I ^ 3 − I ¯ I ^ 2 + σ ( I ^ − D N ) = 0
    Eq. (1)

    where I ^ and − I represent the sought value and local mean, respectively, DN denotes the input value of the digital number, and σ denotes the original image variance. Wiener filter, known as the least mean square filter, was proposed in the 1940s, and it has the ability to restore images even when they are corrupted or blurred. The Wiener filter computes the local image variance. Consequently, it applies less smoothing when the local variance of the image is large and more smoothing when the local variance is small. This approach of Wiener filter (gwiener) provides results superior to those of linear filtering[26]. The mathematical formula for the Wiener filter is as follows:

    g w i e n e r = m + σ 2 − ν 2 σ 2 ( f ( i ) − m )
    Eq. (2)

    where m and σ represent the mean and standard deviation values of the signal within the window, respectively, and ν and f(i) represent the standard deviation and pixel values of noise, respectively. By substituting the mean value in the Wiener filter with the median value, the MMWF is formulated to enhance both resolution and image quality[27,28]. The MMWF is represented by replacing the mean value (m) of the Wiener filter in Eq. (2), with a median value ( m ^ ), as shown in Eq. (3).

    g m m w f = m ^ + σ 2 − ν 2 σ 2 ( f ( i ) − m ^ )
    Eq. (3)

    The Lee filter, introduced in 1981, is based on a multiplicative speckle model and utilizes local statistics to preserve intricate features. The Lee filter operates based on variance; in other words, it applies smoothing to areas with low variance but not to those with high variance. This adaptability allows the Lee filter to preserve details at both low and high contrast[29]. The result of Lee filtering for the smoothed pixel’s gray level value R is as follows:

    R = I c ∗ W + I m ∗ ( 1 − W )
    Eq. (4)

    where Ic and Im are the central pixels of the filter window and the mean intensity within the filter window, respectively; W is ( 1 − C n 2 / C i 2 ) , where Cn and Ci are 1 / ∈ and S/Im, respectively, and S is the standard deviation of the intensity within the filter window. Weighting function W is a measure of estimated noise variation coefficient Cn with respect to image variation coefficient Ci . The number of parameters, ∈ is used to estimate the noise variance and it controls the amount of smoothing applied to the image by the filter.

    3. Quantitative evaluation of noise level

    To quantitatively evaluate the speckle noise reduction ability of various filters with increasing window size in the blood vessel region of liver ultrasound images with various noise intensities, we a set background and a region of interest (ROI) with sizes of 20×20, respectively, and measured the COV and CNR (Fig. 1)[30,31]. Moreover, based on COV and CNR results, we conducted comparative evaluation with the filter that showed the most superior value in quantitative evaluation and the noisy image without the filter to evaluate speckle noise reduction ability.

    The COV　was calculated to compare the variances in the obtained liver ultrasound images.

    C O V = σ R μ
    Eq. (5)

    C N R = | S R − S B K | σ R 2 + σ B K 2
    Eq. (6)

    where σR and σBK represent the standard deviation of signal intensity in the ROI and background, respectively, μ is the average value in the ROI, and SR and SBK denote the average value of signal intensity in the ROI and background, respectively.

    4. Visual evaluation

    We compared the filtered images with a window size of 7×7 with the noisy images. A visual evaluation was performed by enlarging the region indicated by the red box in Fig. 1. The visual assessment criteria were based on how effectively noise in tissue and blood vessel regions were removed and how accurately the signal of blood vessel walls with high pixel values was restored without distortion.

    Ⅲ. Results

    1. Quantitative evaluation

    A quantitative evaluation showed that the COV values improved as the window size of the noise reduction filters increased at noise intensities of 0.01, 0.05, 0.10, and 0.50. Moreover, the Lee filter showed the lowest values for COV results, followed by the Wiener filter, MMWF, and gamma filter (Fig. 2). Additionally, we confirmed that the CNR values improved with increasing window size at various noise intensities, with the Lee filter performing the highest values, followed by the Wiener filter, MMWF, and gamma filters (Fig. 3).

    Based on the results of these analyses, we derived the Lee filter as the most effective filter for speckle noise reduction. To evaluate its usefulness, we compared the COV and CNR values of the noisy image without the noise reduction filter and the Lee filter image with a window size of 7×7 (Fig. 4). The quantitative evaluation showed that at noise intensities of 0.01, 0.05, 0.10, and 0.50, the COV values of the noisy images were 0.30, 0.36, 0.43, and 0.77, respectively, and the CNR values were 1.85, 7.35, 0.93, and 0.57, respectively. Moreover, the COV and CNR values of the Lee filter with a window size of 7×7 showed COV values of 0.23, 0.23, 0.24, and 0.24, and CNR values of 4.06, 4.00, 3.64, and 3.16 at the respective noise intensities.

    The comparative evaluation showed that the Lee filter image at noise intensities of 0.05, 0.01, 0.10, and 0.50 improved the COV by approximately 1.28, 1.51, 1.88, and 3.38 times, respectively, and the CNR values by approximately 2.18, 2.96, 3.91, and 5.50 times, respectively.

    2. Visual evaluation

    We performed the visual evaluation with a window size of 7×7 and the results are shown in Fig. 5. The visual evaluation showed that the Wiener filter and MMWF images were not effective in reducing the speckle noise in the blood vessel wall where the signal was strong, and the speckle noise in the liver tissue region intensified as the noise intensity increased. In addition, the gamma filter image exhibited a decrease in the overall signal value, thereby decreasing the contrast and brightness of the image. For the Lee filter image, the speckle noise was effectively reduced at all noise intensities; however, blurring was also considered, resulting in a decrease in resolution. Nonetheless, the Lee filter effectively reduced speckle noise at all noise intensities and there was no variation in contrast and brightness that occurred with the gamma filter. When comparing images with the Lee filter applied by noise intensity, we found that the smaller the noise intensity, the more effective the noise reduction, however, at a noise intensity of 0.50, speckle noise was reduced but a grid artifact was introduced.

    Ⅳ. Discussion

    Liver ultrasound is a widely used diagnostic method due to its real-time imaging capabilities, non-invasive nature, and absence of radiation or contrast agents [9,32,33]. However, ultrasound imaging diagnostic procedures have the disadvantage owing to the speckle noise[18,19]. To solve these problems, various techniques for speckle noise reduction have been proposed[25-29]. Although previous research was conducted by improving the hardware, thereby significantly impacting the ultrasound image quality, technical limitations such as economic difficulties and complexity existed. Hence, research was actively conducted by improving the software[21,22]. One software technique, filtering, is the most widely used because it can be applied to existing ultrasound equipment, is simple to modify, and is economically beneficial for simple maintenance[34].

    A gamma filter, a widely used nonlinear filter for speckle noise reduction, can effectively reduce noise by recognizing the features of the input data. It has the advantage of correcting the spectrum by assigning higher weights to regions with large variations in the input data. However, this characteristic does not solve the problem of computational cost. Moreover, applying this to certain data is difficult because this requires assumptions about the distribution of the input data[25]. The Wiener filter, which is known to efficiently reduce speckle noise, is a common linear filter used for speckle noise reduction. The Wiener filter calculates the power spectral density of the input data and the average value of the spectral density to calculate the signal to noise ratio (SNR) for estimating the spectrum, and it uses the estimated spectrum to restore the image. However, the Wiener filter works most effectively when the noise distribution is Gaussian. However, the speckle noise generated in real ultrasound images is often irregular and nonlinear in distribution; therefore, it does not perform effectively[26]. To address the limitations of the Wiener filter, the median modified Wiener filter (MMWF) is proposed. The MMWF uses spatial and time correlations estimated from multiple frames to reduce the speckle noise. The MMWF is effective when noise has different intensities and directions. Moreover, the MMWF can effectively reduce speckle noise when the noise does not have a Gaussian distribution. However, the MMWF is highly computationally intensive and has many variables; therefore, the filtering result may depend on the setting of the variables[27,28].

    To address these disadvantages, the Lee filter can be used because it is less computationally intensive and requires no assumptions about the input data; it can be applied to a wide variety of data. In addition, the Lee filter functions effectively when the input data are not uniform, because it considers the uniformity of the input data and the statistical information of the neighboring pixels to perform filtering. Compared with conventional filters, the Lee filter has the advantage of more accurately preserving information near the edge, enabling a detailed depiction of the anatomical structures[29].

    In the case of liver disease, liver function depends on the inflow and regulation of blood; therefore, the ability to differentiate between tissues and blood vessels when performing an ultrasound scan is crucial to assess the blood supply to the liver. Specifically, liver veins are critical for diagnosing liver diseases such as liver cancer that can be difficult to detect at an early stage[15,16]. However, whereas many studies have been conducted to reduce speckle noise in liver ultrasound images, relatively few studies have been conducted on reducing noise in the blood vessel region of the liver[35,36]. Moreover, the intensity of speckle noise in ultrasound examinations is affected by the type of ultrasound equipment, survey location, and patient physiology, and the intensity of speckle noise in these clinical settings is different; however, research on the application of filters to images with various intensities of speckle noise is lacking[37,38]. Therefore, in this study, the noise intensity was varied to evaluate appropriate noise reduction techniques in various clinical environments. Furthermore, various speckle noise reduction filters were applied to reduce the speckle noise in the blood vessel region of the liver from the acquired noisy images to quantitatively evaluate the noise reduction performance of those filters.

    A quantitative evaluation showed that the noise reduction performance of the speckle noise reduction filters improved as the window size increased. For noise intensities of 0.01, 0.05, 0.10, and 0.50, we confirmed that the Lee filter exhibited the most superior noise reduction performance, followed by the Wiener filter, MMWF, and gamma filter. The research findings are attributed to the Lee filter's ability to effectively reduce speckle noise by taking into account the statistical relationships among neighboring pixels, resulting in a better match with the distribution of neighboring pixel values. Moreover, when comparing the noisy image with the image of the Lee filter with a window size of 7×7 that showed the most improved noise reduction ability, the COV and CNR values improved significantly as the speckle noise intensity increased. The COV increased from a minimum of 0.23 to a maximum of 0.24, and the CNR increased from a minimum of 3.16 to a maximum of 4.06.

    As a result, this study demonstrated the usefulness of the Lee filter in liver ultrasound images acquired in various clinical settings through studies performed at various noise intensities and is expected to serve as a basis for ultrasound studies of liver vessels that have not been sufficiently studied. Additionally, the application of the Lee filter in the process of observing the blood vessels of the liver that are relatively difficult to detect because of their small size and complex structure, such as the hepatic artery compared to the hepatic vein, is expected to contribute to the observation of blood vessel areas and increase the visibility of small structures by effectively reducing noise[39].

    A limitation of this study is that the speckle denoising filters used in the experiments were applied to observe only the reduction in the amount of noise; hence, blurring and resolution degradation could not be evaluated. Additionally, a limitation of this study is that it conducted experiments using a single liver ultrasound image. Therefore, further research is required on various image quality evaluations, except for noise reduction performance and planned to validate the utility of the Lee filter for speckle noise reduction by acquiring additional liver ultrasound images. Moreover, we have plans for future research that will involve the comparative evaluation of filters across diverse liver ultrasound images based on varying window sizes. This research will also involves studies aimed at optimizing the window size.

    Nevertheless, our study offer clinical application possibility. The incorporation of Lee filters into standard liver ultrasound image processing procedures offers healthcare practitioners a valuable tool for improving the clarity and diagnostic value of ultrasound scans. This advantage is further emphasized by the flexibility of Lee filters, allowing for patient-specific adjustments in window size. By aligning filter settings with individual patient characteristics and the clinical environment, optimal speckle noise reduction is achieved. Therefore, our study provide that implementation of this approach in routine clinical settings holds the potential to enhance diagnostic precision, ultimately contributing to improved patient outcomes. While our study currently emphasizes speckle noise reduction in the liver blood vessel region, it holds the potential for broader applications in liver ultrasound imaging. Extending these noise reduction approaches to other liver regions, including the parenchyma, biliary ducts, and lesions, promises enhanced overall image quality and diagnostic capabilities. This enhancement in ultrasound image quality, achieved through speckle noise reduction, supports accurate diagnoses, treatment planning, and underscores the significance of early detection in liver disease.

    Additionally, our study suggests future possibilities for the advancement of speckle noise reduction techniques, which could lead to further research and application programs for better patient outcomes and disease management. In particular, in the rapidly developing field of deep learning, the application of the Lee filter to dataset construction is expected to provide accurate and generalized results through efficient speckle noise reduction[40]. This approach has the potential to improve the performance and efficiency of deep learning models by applying high-quality training data.

    Ⅴ. Conclusions

    In this study, to reduce the speckle noise in the liver blood vessel region, we modeled and used speckle noise reduction filters to reduce speckle noise in clinical liver ultrasound images. To simulate various clinical conditions, we added various noise intensities. Additionally, we performed a quantitative analysis by comparing the Lee filter with various filtering techniques and window size when applied to noisy images, confirming the speckle noise ability of the Lee filter. In the experimental results, we observed significant improvements across all quantitative evaluation factors with the application of the Lee filtering technique, surpassing the performances of various speckle noise reduction filters. These results suggest that the Lee filter can replace other speckle noise filters, and it is an effective noise reduction filter in the liver blood vessel region.

    Figure

    JRST-46-6-475_F1.gif

    An acquired clinical liver ultrasound image with each rectangular-shaped background (BK) and region of interest (ROI) for quantitative evaluation.

    JRST-46-6-475_F2.gif

    Results of coefficient of variation (COV) according to the ROI shown in Fig. 1 in a clinical liver ultrasound image with the Wiener filter, MMWF, Gamma filter, and Lee filter with various speckle noise intensiti: (esa) 0.01, (b) 0.05, (c) 0.10, and (d) 0.50, respectively.

    JRST-46-6-475_F3.gif

    Results of contrast to noise ratio (CNR) according to the ROI shown in Fig. 1 in a clinical liver ultrasound image with the Wiener filter, MMWF, Gamma filter, and Lee filter with various speckle noise intensities: (a) 0.01, (b) 0.05, (c) 0.10, and (d) 0.50, respectively.

    JRST-46-6-475_F4.gif

    Comparative evaluation results (a) COV and (b) CNR of the Lee filter with a 7×7 window size and noisy images with various noise intensities.

    JRST-46-6-475_F5.gif

    Comparative visual evaluation of various filters with a 7×7 window size and noisy image with noise intensities of 0.01, 0.05, 0.10, and 0.50.

    Table

    Reference

    1. World Health Organization. The top 10 causes of death. [cited 2023 Sep 4]. Retrieved from https://www.wh o.int/news-room/fact-sheets/detail/the-top-10- causes-of-death
    2. Statistics Korea. Causes of death statistics in 2019. [cited 2023 Sep 4]. Retrieved from https://kostat.g o.kr/board.es?mid=a10301060200&bid=218&act=vi ew&list_no=385219
    3. Brancatelli G, Federle MP, Vilgrain V, Vullierme MP, Marin D, Lagalla R. Fibropolycystic liver disease: CT and MR imaging findings. RadioGraphics. 2005;25(3):659-70.
    4. Boll DT, Merkle EM. Diffuse liver disease: Strategies for hepatic CT and MR imaging. RadioGraphics. 2009;29(6):1591-614.
    5. Ballestri S, Romagnoli D, Nascimbeni F, Francica G, Lonardo A. Role of ultrasound in the diagnosis and treatment of nonalcoholic fatty liver disease and its complications. Expert Rev. Gastroenterol. Hepatol. 2015;9(5):603-27.
    6. Gerstenmaier JF, Gibson RN. Ultrasound in chronic liver disease. Insights Imaging. 2014;5:441-55.
    7. Goyal N, Jain N, Rachapalli V, Cochlin DL, Robinson M. Non-invasive evaluation of liver cirrhosis using ultrasound. Clin. Radiol. 2009;64(11):1056–66.
    8. Solbiati L, Ierace T, Tonolini M, Cova L. Guidance and monitoring of radiofrequency liver tumor ablation with contrast-enhanced ultrasound. Eur. J. Radiol. 2004;51:S19-23.
    9. Lupşor-Platon ML, Stefănescu H, Mureșan D, Florea M, Szász ME, Maniu A, et al. Noninvasive assessment of liver steatosis using ultrasound methods. Med. Ultrason. 2014;16(3):236-45.
    10. Schwenzer NF, Springer F, Schraml C, Stefan N, Machann J, Schick F. Non-invasive assessment and quantification of liver steatosis by ultrasound, computed tomography and magnetic resonance. J. Hepatol. 2009;51(3):433–45.
    11. Acharya UR, Raghavendra U, Fujita H, Hagiwara Y, Koh JE, Jen Hong T, et al. Automated characterization of fatty liver disease and cirrhosis using curvelet transform and entropy features extracted from ultrasound images. Comput. Biol. Med. 2016;79: 250-8.
    12. Nicolau C, Brú C, Carreras E, Bosch J, Bianchi L, Gilabert R, et al. Sonographic diagnosis and hemodynamic correlation in veno-occlusive disease of the liver. J. Ultrasound Med. 1993;12(8):437–40.
    13. Sagi R, Reif S, Neuman G, Webb M, Phillip M, Shalitin S. Nonalcoholic fatty liver disease in overweight children and adolescents. Acta Paediatr. 2007;96(8):1209-13.
    14. Amitrano L, Guardascione MA, Brancaccio V, Margaglione M, Manguso F, Iannaccone L, et al. Risk factors and clinical presentation of portal vein thrombosis in patients with liver cirrhosis. J. Hepatol. 2004;40(5):736-41.
    15. Mwanza T, Miyamoto T, Okumura M, Hagio M, Fujinaga T. Ultrasonography and angiographic examination of normal canine liver vessels. Jpn. J. Vet. Res. 1996;44(3):179-88.
    16. Zhou, JH, Li AH, Cao LH, Jiang HH, Liu LZ, Pei XQ, et al. Haemodynamic parameters of the hepatic artery and vein can detect liver metastases: Assessment using contrast-enhanced ultrasound. Br. J. Radiol. 2008;81(962):113-9.
    17. Marzouni HZ, Davachi B, Rezazadeh M, Milani MS, Matinfard S. Diagnostic value of hepatic vein ultrasound in early detection of liver cirrhosis. Galen Med. J. 2018;7:e1140.
    18. Jeyalakshmi TR, Ramar K. A modified method for speckle noise removal in ultrasound medical images. Int. J. Comput. Electr. Eng. 2010;2(1):54.
    19. Karthikeyan K, Chandrasekar C. Speckle noise reduction of medical ultrasound images using bayesshrink wavelet threshold. Int. J. Comput. Appl. 2011; 22(9):8-14.
    20. Krissian K, Kikinis R, Westin CF, Vosburgh K. Speckle-constrained of ultrasound images. 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition(CVPR' '05). 2005; 2:547-52.
    21. Singh K, Ranade SK, Singh C. A hybrid algorithm for speckle noise reduction of ultrasound images. Comput. Method. Programs Biomed. 2017;148:55-69.
    22. Outtas M, Serir A, Kerouh F. Speckle noise reduction in ultrasound image based on A Multiplicative Multiresolution Decomposition (MMD). Eighth ed. of Int. Symp. On Sugnal, Image, Video and Communications; 2014.
    23. Leal AS, Paiva HM. A new wavelet family for speckle noise reduction in medical ultrasound images. Measurement. 2019;140:572-81.
    24. Podilchuk C, Bajor M, Stoddart W, Barinov L, Hulbert W, Jairaj A, et al. Speckle reduction using stepped-frequency continuous wave ultrasound. 2012 IEEE Signal Processing in Medicine and Biology Symposium(SPMB). 2012:1-4.
    25. Lopes A, Touzi R, Nezry E. Adaptive speckle filters and scene heterogeneity. IEEE Trans. Geosci. Remote Sensing. 1990;28(6):992-1000.
    26. Sudha S, Suresh GR, Sukanesh R. Speckle noise reduction in ultrasound images by wavelet thresholding based on weighted variance. Int. J. Comput. Theor. Eng. 2009;1(1):7.
    27. Cannistraci CV, Abbas A, Gao X. Median modified wiener filter for nonlinear adaptive spatial denoising of protein NMR multidimensional spectra. Sci. Rep. 2015;5(1):8017.
    28. Park CR, Kang S, Lee Y. Median modified wiener filter for improving the image quality of gamma camera images. Nucl. Eng. Technol. 2020;52(10): 2328-33.
    29. Lee JS. Digital image enhancement and noise filtering by use of local statistics. IEEE Trans. Pattern Anal. Mach. Intell. 1980;2:165-8.
    30. Chen J, Benesty J, Huang Y, Doclo S. New insights into the noise reduction Wiener filter. IEEE Trans. Aud. Speech Lang. Process. 2006;14(4):1218-34.
    31. Rodriguez-Molares A, Rindal OMH, D'hooge J, Masoy SE, Austeng A, Lediju Bell MA, et al. The generalized contrast-to-noise ratio: A formal definition for lesion detectability. IEEE Trans. Ultrason. Ferroelectr. Freq. Control. 2019;67(4): 745–59.
    32. Cho JY, Ye SY. GLCM algorithm image analysis of nonalcoholic fatty liver and focal fat sparing zone in the ultrasonography. Journal of Radiological Science and Technology. 2017;40(2):205-11.
    33. Kim DH, Kwon DM. Performance testing of medical US equipment using US phantom(ATS-539)(Focusing on Daegu Region). Journal of Radiological Science and Technology. 2014;37(4):295-305.
    34. Khan MN, Altalbe A. Experimental evaluation of filters used for removing speckle noise and enhancing ultrasound image quality. Biomed. Signal Process. Control. 2022;73:103399.
    35. Jabarulla MY, Lee HN. Speckle reduction in ultrasound liver image based on sparse representation over a learned dictionary. Applied Sciences. 2018; 8(6):903.
    36. Ramamoorthy S, Siva Subramanian R, Gandhi D. An efficient method for speckle noise reduction in ultrasound liver images for e-health applications. Distributed Computing and Internet Technology: 10th International Conference, ICDCIT. 2014:311-21.
    37. Pregitha RE, Jagathesan V, Selvakumar CE. Speckle noise reduction in ultrasound fetal images using edge preserving adaptive shock filters. Int. J. Sci. Res. Publ. 2012;2(3):1-3.
    38. Islam MA, Talukder MH, Hasan MM. Speckle noise reduction from ultrasound image using modified binning method and fuzzy inference system. 2013 2nd International Conference on Advances in Electrical Engineering(ICAEE). 2013:359-62.
    39. Loupas T, McDicken WN, Allan PL. Noise reduction in ultrasonic images by digital filtering. Br. J. Radiol. 1987;60(712):389-92.
    40. Gupta S, Gupta A. Dealing with noise problem in machine learning data-sets: A systematic review. Procedia Comput. Sci. 2019;161:466-74.
    Copyright © Korean Society of Radiological Science. All rights reserved.