Original contribution
Breast ultrasound image enhancement using fuzzy logic

https://doi.org/10.1016/j.ultrasmedbio.2005.10.007Get rights and content

Abstract

Breast cancer is still a serious disease in the world. Early detection is very essential for breast cancer prevention and diagnosis. Breast ultrasound (US) imaging has been proven to be a valuable adjunct to mammography in the detection and classification of breast lesions. Because of the fuzzy and noisy nature of the US images and the low contrast between the breast cancer and tissue, it is difficult to provide an accurate and effective diagnosis. This paper presents a novel algorithm based on fuzzy logic that uses both the global and local information and has the ability to enhance the fine details of the US images while avoiding noise amplification and overenhancement. We normalize the images and then fuzzify the normalized images based on the maximum entropy principle. Edge and textural information are extracted to describe the lesion features and the scattering phenomenon of US images and the contrast ratio measuring the degree of enhancement is computed and modified. The defuzzification process is used to obtain the enhanced US images. To demonstrate the performance of the proposed approach, the algorithm was tested on 86 breast US images. Experimental results confirm that the proposed method can effectively enhance the details of the breast lesions without overenhancement or underenhancement. (E-mail: [email protected])

Introduction

Breast cancer is still one of the most common cancers and a serious disease among women in the world. Because the causes of this disease still remain unknown, early detection is essential for diagnosis and it makes the treatment more effective in reducing the mortality and cost.

Mammography and ultrasonography have been preferred methods for breast cancer detection. Mammography is very sensitive, but not specific, for detecting breast cancer (Joseph et al. 1999). As a result, approximately 65% of the cases referred for surgical biopsy are actually benign lesions (Kopans 1992, Knutzen and Gisvold 1993). Mammography also has limitations for cancer detection in dense breast tissue of young patients. Currently, breast ultrasound (US) imaging is a valuable adjunct to mammography in the early detection and classification of breast lesions (Drukker et al. 2004). Sonography was more effective for women younger than 35 years old (Bassett et al. 1991). The results (Laine et al. 1995) suggest that the more dense the breast parenchyma, the higher is the detection accuracy of malignant tumors using US. The accuracy rate of breast US imaging has been reported to be 96 to 100% in the diagnosis of simple benign cysts (Jackson 1990). Breast US examination has played a more and more significant role in detecting breast cancers because of the fact that sonography has the ability to show masses obscured mammographically by dense tissue and has no ionizing radiation and because of its low cost and portability (Drukker et al. 2004).

Image enhancement is used to improve the quality of the image and to correct deficiencies of the contrast. Many algorithms have been proposed to enhance images. An automatic image-enhancement algorithm driven by an evolutionary optimization process was studied (Munteanu and Rosa 2004) that proposed an objective criterion and used an evolutionary algorithm as a global search strategy for the best enhancement. However, the evolutionary algorithm requires making a series of trial enhancements to find the final result and the algorithm has quite high computational complexity. That paper only compared the results with those of classical linear contrast stretching and histogram equalization techniques that are quite old. One method generalized the linear scale spaces in the complex domain by combining the diffusion equation with the free Schrödinger equation for image enhancement and denoising (Gilboa et al. 2004). Two examples of nonlinear complex processes were developed; a regularized shock filter for image enhancement and a ramp-preserving denoising process. However, they did not discuss underenhancement and overenhancement. Three quantitative measures of contrast enhancement of mammographic images in a computer-aided system were proposed (Singh and Bovis 2005) that can be used for selecting the best-suited image enhancement on a per mammogram basis. However, the method was only tested on six enhancement algorithms.

It is well known that breast US images have some degree of fuzziness, such as indistinct cyst borders, ill-defined mass shapes and different tumor densities. Some nonlinear filters were used to enhance the US images. A morphologic method, alternating sequential filter, to enhance US images that is an iterative application of openings and closings with structuring elements of increasing sizes, was studied (Tsubai et al. 2000). No objective method for determining the number of iterations was supplied and it is difficult to enhance the image and remove noise at the same time. A nonlinear enhancing filter is based on sorting the elements in a moving window and extracting statistical characteristics from them (Lee et al. 2001). The filter compared the statistical values of the front and back points with those of the center point in the window and estimated the output using the compared result. However, the enhancement was affected by the size of the filters and how to determine the size and number of filters was not discussed in detail. Many nonlinear or linear map functions are used to enhance the contrast of US image in space or other domains. The grey-level mapping technique was applied to enhance the US images having different levels of contrast and brightness (Saim et al. 2000). Grey-level mapping is a technique to map the input grey levels (low and high) to the stretched output grey levels (bottom and top) observed in a look-up table. The mapping function is an exponential function and its parameter is a constant that cannot vary with different images. A nonlinear algorithm was studied for contrast enhancement, which is accomplished via nonlinear stretching followed by hard thresholding of wavelet coefficients within midrange spatial frequency levels (Zhou et al. 2002). But the selection of the threshold is subjective and the image enhancement is highly dependent on the value of the threshold.

An algorithm uses line segments (called “sticks”) in different angular orientations as templates and selects the orientation at each point that is most likely to represent a line in the image to improve edge information, making the line segments more suitable for edge detection and for US image enhancement (Pathak et al 2000, Pathak and Kim 2000, Abolmaesumi and Sirouspour 2004, Awad et al 2003). However, the algorithm only enhances the edge information and the features inside the tissues and lesions are not affected.

Fuzzy set theory is used to enhance US images (Li et al 2003, Li et al 2004). The image is transformed into fuzzy domain using membership function and then the membership is enhanced by an iterative four-segment function. However, the rules of the iterative four-segment function and the iteration time are fixed and subjective. Also, the method only uses global features that cannot reflect the local contrast change. A new fuzzy logic filter for image enhancement was reported (Farbiz et al. 2000) that was based on fuzzy-logic control, with the ability to remove impulsive noise and smooth Gaussian noise while simultaneously preserving edges and image details. However, how to deal with overenhancement and underenhancement was not discussed.

A statistical model using tissue properties and intensity inhomogeneities in US was presented and used in contrast enhancement and image segmentation and the maximum a posteriori principle was used to correct tissue intensity and to conduct contrast enhancement of breast US images (Xiao et al. 2000). The methods only modified the distribution of the intensities and did not pay much attention to the features of tumors and tissues.

Among the early indicators of breast cancers, masses and microcalcifications are the primary features (Lanyi 1986) that are very important visual features of early cancers found by radiologists (Mclelland 1990). Unfortunately, at the early stage of breast cancer, the inside structure and border of masses of US images are very subtle and varied in appearance; hence, it makes the diagnosis very difficult. The difference between the suspicious areas and normal tissues can be very slight. Breast US image enhancement, especially for the images of dense breasts, is exceedingly important for both the doctors and computer-aided diagnosis systems. After enhancement, more useful information can be extracted.

In this paper, we propose a novel contrast-enhancement algorithm based on fuzzy logic and the characteristics of breast US images. The maximum fuzzy entropy principle is used to map the original image, then the characteristics of the US image are taken into account. Specifically, the edge and textural information is extracted to evaluate the lesion features and the scattering phenomenon of US images and the local information is used to define the enhancement criterion. Finally, the algorithm enhances the details and lesion features using the local fuzzy information.

The paper is organized as follows. In the next section, the proposed method is presented in detail. Experiments are then discussed to demonstrate the performance of the algorithm. Finally, the conclusions are drawn.

Section snippets

Proposed approach

The proposed method consists of the following steps; image normalization, image fuzzification, edge information extraction, textural information extraction and contrast enhancement.

Experimental results

The breast US images used in the experiments were provided by the Second Affiliated Hospital of Harbin Medical University, Harbin, China. The images were collected by using a Vivid 7 (GE, Horten, Norway) with a 5- to 14-MHz linear probe and were captured directly from the video signals. The database consisted of a total of 86 images of 49 cases and each single lesion is in one image. Of the 49 cases, 14 were benign solid lesions (30 images) and 35 were malignant solid lesions (56 images). For

Conclusions

A breast US image-enhancement algorithm based on fuzzy logic and fuzzy entropy was developed. From the experiments, the proposed algorithm makes the details of ROIs much clearer and it has no overenhancement or underenhancement. The good performance is caused by the following factors. 1 The S function was used in image fuzzification and the parameters were determined using the maximum entropy principle. Based on the information theory, the fuzzified images contain maximum information. 2. The

References (38)

  • F. Farbiz et al.

    A new fuzzy logic filter for image enhancement

    IEEE Trans Syst Man Cybernet Part B

    (2000)
  • G. Gilboa et al.

    Image enhancement and denoising by complex diffusion processes

    IEEE Trans Patt Anal Machine Intell

    (2004)
  • R.C. Gonzalez et al.

    Digital image processing

    (2002)
  • E.L. Hall

    Computer image processing and recognition

    (1979)
  • V.P. Jackson

    The role of ultrasound in breast imaging

    Radiology

    (1990)
  • Y.L. Joseph et al.

    Application of artificial neural networks for diagnosis of breast cancer

    Proc Cong Evolutionary Comput

    (1999)
  • D.B. Kopans

    The positive predictive value of mammography

    Am J Roentgenol

    (1992)
  • M. Lanyi

    Diagnosis of breast calcifications

    (1986)
  • K.I. Laws

    Texture energy measures

  • Cited by (0)

    View full text