Limits...
Shape-based Automatic Detection of Pectoral Muscle Boundary in Mammograms.

Chen C, Liu G, Wang J, Sudlow G - J Med Biol Eng (2015)

Bottom Line: The seed point is then detected on the initial boundary and the pectoral boundary is evolved from candidate points produced using a shape-based growth strategy.A cubic polynomial fitting function is implemented to obtain the final pectoral muscle boundary.A 97.2 % acceptable rate from expert radiologists and assessment results based on the false positive rate, false negative rate, and Hausdorff distance demonstrate the robustness and effectiveness of the proposed shape-based detection method.

View Article: PubMed Central - PubMed

Affiliation: Department of Biomedical Engineering, Nanjing University of Aeronautics and Astronautics, Nanjing, 210016 China.

ABSTRACT

The detection of the pectoral muscle boundary in the medio-lateral oblique view of mammograms is essential to improving the computer-aided diagnosis of breast cancer. In this study, a shape-based detection method is proposed for accurately extracting the boundary of the pectoral muscle in mammograms. A shape-based enhancement mask is applied to the mammogram and the initial boundary is then defined using morphological operators. The seed point is then detected on the initial boundary and the pectoral boundary is evolved from candidate points produced using a shape-based growth strategy. A cubic polynomial fitting function is implemented to obtain the final pectoral muscle boundary. The proposed method was applied to 322 mammograms from the mini Mammographic Image Analysis Society database. A 97.2 % acceptable rate from expert radiologists and assessment results based on the false positive rate, false negative rate, and Hausdorff distance demonstrate the robustness and effectiveness of the proposed shape-based detection method.

No MeSH data available.


Related in: MedlinePlus

Linear enhancement mask
© Copyright Policy - OpenAccess
Related In: Results  -  Collection


getmorefigures.php?uid=PMC4491117&req=5

Fig2: Linear enhancement mask

Mentions: The pectoral muscle usually has an approximate direction and higher gray level intensity in mammogram. Based on the prior knowledge, an enhancement filter is commonly used to process mammograms to highlight the pectoral muscle. Zhou et al. [12] developed a gradient-based directional kernel (GDK) filter to enhance the linear texture structures on mammograms at approximately 45° from the top left to the bottom right and implemented it by convolving the image with an 11 × 11 mask with values of 1 and −1. However, the GDK filter is very sensitive to the ridge points and produces a lot of unwanted boundaries. To overcome these problems, a linear shape-based enhanced filter with several coefficients that considers the transition intensity changes around the pectoral muscle edge is proposed here. The filter is designed as:1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$g(x,y) = w_{s,t} \sum\limits_{s = 0}^{M} {\sum\limits_{t = 1}^{N} {(I(x + s,y - t) - I(x + s,y + t)) + w_{c} I(x,y)} }$$\end{document}g(x,y)=ws,t∑s=0M∑t=1N(I(x+s,y-t)-I(x+s,y+t))+wcI(x,y)where is the intensity of the pixel at point (x, y), (M + 1) is the number of rows, is the number of pixel pairs contributing to the weighted differentiation along the horizontal direction, and and are weight coefficients. The expression can be implemented by convolving a mammogram with a linear enhancement mask (Fig. 2). Considering that the pectoral muscle gradually narrows from top to bottom, the bottom coefficients of the mask are shifted to the left to highlight the structural characteristics, as shown in Fig. 3. Due to the pixel intensity gradually becoming stronger away from the left side of the boundary, in this mask, the coefficient increases with to suppress unwanted tissues, and the diagonal coefficients enhance the structural characteristics of the pectoral muscle. The sum of the mask coefficients (excluding ) is zero. In the filtered image, the regions with homogenous intensities in the original image are suppressed and the boundary is emphasized. The coefficient , which represents the contribution of the center point, is often set in the range of 0–1 to avoid excessive influence on filter results while processing non-boundary regions with a high intensity value.Fig. 2


Shape-based Automatic Detection of Pectoral Muscle Boundary in Mammograms.

Chen C, Liu G, Wang J, Sudlow G - J Med Biol Eng (2015)

Linear enhancement mask
© Copyright Policy - OpenAccess
Related In: Results  -  Collection

Show All Figures
getmorefigures.php?uid=PMC4491117&req=5

Fig2: Linear enhancement mask
Mentions: The pectoral muscle usually has an approximate direction and higher gray level intensity in mammogram. Based on the prior knowledge, an enhancement filter is commonly used to process mammograms to highlight the pectoral muscle. Zhou et al. [12] developed a gradient-based directional kernel (GDK) filter to enhance the linear texture structures on mammograms at approximately 45° from the top left to the bottom right and implemented it by convolving the image with an 11 × 11 mask with values of 1 and −1. However, the GDK filter is very sensitive to the ridge points and produces a lot of unwanted boundaries. To overcome these problems, a linear shape-based enhanced filter with several coefficients that considers the transition intensity changes around the pectoral muscle edge is proposed here. The filter is designed as:1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$g(x,y) = w_{s,t} \sum\limits_{s = 0}^{M} {\sum\limits_{t = 1}^{N} {(I(x + s,y - t) - I(x + s,y + t)) + w_{c} I(x,y)} }$$\end{document}g(x,y)=ws,t∑s=0M∑t=1N(I(x+s,y-t)-I(x+s,y+t))+wcI(x,y)where is the intensity of the pixel at point (x, y), (M + 1) is the number of rows, is the number of pixel pairs contributing to the weighted differentiation along the horizontal direction, and and are weight coefficients. The expression can be implemented by convolving a mammogram with a linear enhancement mask (Fig. 2). Considering that the pectoral muscle gradually narrows from top to bottom, the bottom coefficients of the mask are shifted to the left to highlight the structural characteristics, as shown in Fig. 3. Due to the pixel intensity gradually becoming stronger away from the left side of the boundary, in this mask, the coefficient increases with to suppress unwanted tissues, and the diagonal coefficients enhance the structural characteristics of the pectoral muscle. The sum of the mask coefficients (excluding ) is zero. In the filtered image, the regions with homogenous intensities in the original image are suppressed and the boundary is emphasized. The coefficient , which represents the contribution of the center point, is often set in the range of 0–1 to avoid excessive influence on filter results while processing non-boundary regions with a high intensity value.Fig. 2

Bottom Line: The seed point is then detected on the initial boundary and the pectoral boundary is evolved from candidate points produced using a shape-based growth strategy.A cubic polynomial fitting function is implemented to obtain the final pectoral muscle boundary.A 97.2 % acceptable rate from expert radiologists and assessment results based on the false positive rate, false negative rate, and Hausdorff distance demonstrate the robustness and effectiveness of the proposed shape-based detection method.

View Article: PubMed Central - PubMed

Affiliation: Department of Biomedical Engineering, Nanjing University of Aeronautics and Astronautics, Nanjing, 210016 China.

ABSTRACT

The detection of the pectoral muscle boundary in the medio-lateral oblique view of mammograms is essential to improving the computer-aided diagnosis of breast cancer. In this study, a shape-based detection method is proposed for accurately extracting the boundary of the pectoral muscle in mammograms. A shape-based enhancement mask is applied to the mammogram and the initial boundary is then defined using morphological operators. The seed point is then detected on the initial boundary and the pectoral boundary is evolved from candidate points produced using a shape-based growth strategy. A cubic polynomial fitting function is implemented to obtain the final pectoral muscle boundary. The proposed method was applied to 322 mammograms from the mini Mammographic Image Analysis Society database. A 97.2 % acceptable rate from expert radiologists and assessment results based on the false positive rate, false negative rate, and Hausdorff distance demonstrate the robustness and effectiveness of the proposed shape-based detection method.

No MeSH data available.


Related in: MedlinePlus