Scholarly article on topic 'Intensity Based Automatic Boundary Identification of Pectoral Muscle in Mammograms'

Intensity Based Automatic Boundary Identification of Pectoral Muscle in Mammograms Academic research paper on "Medical engineering"

Share paper
Academic journal
Procedia Computer Science
OECD Field of science
{Mammogram / "Enhancement filter mask" / "Pectoral muscle" / "Boundary points"}

Abstract of research paper on Medical engineering, author of scientific article — P.S. Vikhe, V.R. Thool

Abstract The pectoral muscle detection is an important assignment to improve the diagnostic performance of the breast cancer detection. In this paper, we have proposed an intensity based approach for pectoral muscle boundary detection in mammograms. The enhancement filter mask of 3×2, have been proposed and applied on the image to enhance the pectoral region of the mammograms. The pectoral boundary points from the candidates were detected based on threshold technique. Finally, all the boundary points detected were connected to obtain the boundary of pectoral muscle. The proposed technique has been tested on 320 digitized mammograms form mini-Mammographic Image Analysis Society (MIAS) database of 322 mammograms, with an acceptance rate of 96.56% from expert radiologists. The mean False Positive (FP) and False Negative (FN) rate demonstrate the effectiveness of the proposed method.

Academic research paper on topic "Intensity Based Automatic Boundary Identification of Pectoral Muscle in Mammograms"


7th International Conference on Communication, Computing and Virtualization 2016

Intensity based Automatic Boundary Identification of Pectoral

Muscle in Mammograms

P. S. Vikhea, V. R. Thoolb

ab S.G.G.S Institute of Engineering and Technology, Vishnupuri, Nanded - 431606, (M.S.) India


The pectoral muscle detection is an important assignment to improve the diagnostic performance of the breast cancer detection. In this paper, we have proposed an intensity based approach for pectoral muscle boundary detection in mammograms. The enhancement filter mask of 3^2, have been proposed and applied on the image to enhance the pectoral region of the mammograms. The pectoral boundary points from the candidates were detected based on threshold technique. Finally, all the boundary points detected were connected to obtain the boundary of pectoral muscle. The proposed technique has been tested on 320 digitized mammograms form mini-Mammographic Image Analysis Society (MIAS) database of 322 mammograms, with an acceptance rate of 96.56% from expert radiologists. The mean False Positive (FP) and False Negative (FN) rate demonstrate the effectiveness of the proposed method.

© 2016 The Authors.Publishedby ElsevierB.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.Org/licenses/by-nc-nd/4.0/).

Peer-review under responsibility of the Organizing Committee of ICCCV 2016 Keywords: Mammogram, Enhancement filter mask, Pectoral muscle, Boundary points.


Available online at


Procedia Computer Science 79 (2016) 262 - 269

1. Introduction

The breast cancer is most common health issue diagnosed, that leads to cause death among women in both developing and developed countries. One out of eight women's in the United States (US) will develop breast cancer at some stage during her life time according to National Cancer Institute [4]. According to World Health Organization (WHO) in 2004, cancer accounted 13% of all deaths in the world [4]. Breast cancer cases and deaths estimated by American Cancer Society (ACS), yearwise of US are mentioned in Table 1 [2].

* Corresponding author. Tel.: +918856000886;

E-mail address:

1877-0509 © 2016 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (

Peer-review under responsibility of the Organizing Committee of ICCCV 2016 doi:10.1016/j.procs.2016.03.034

The radiologist uses X-ray mammography as a screening and diagnosis tool, for detection of breast cancer at preliminary stage. This is the most reliable technique for early detection of breast cancer, reducing the mortality rates up to 25%. Screening mammography is not easy task for radiologists; 10%-30% of lesions are missed during routine screening [3]. The double screening of mammograms can increase the accuracy, but expertise limitations and intra-observer variability limits its use due to huge number of mammograms. The accuracy and speed can be improved using computerized analysis. Also, it reduces the work of radiologists and discrepancies due to inter-observer. The region emerges triangular area across the upper posterior periphery of the image on a proper Medio-Lateral Oblique (MLO) as a high intensity connecting to the chest is called pectoral. Existence of pectoral muscles affects the detection of cancer during computerized processing of mammograms [5].

Table 1. The yearwise estimated cases and deaths of breast cancer.

Year Estimated new cases Estimated deaths

Both Sex Male Female Both Sex Male Female

2005 2,12,930 1,690 2,11,240 40,870 460 40,410

2006 2,14,640 1,720 2,12,920 41,430 460 40,970

2007 1,80,510 2,030 1,78480 40,910 450 40,460

2008 1,84,450 1,990 1,82,460 40,930 450 40,480

2009 1,94,280 1,910 1,92,370 40,610 440 40,170

2010 2,09,060 1,970 2,07,090 40,230 390 39,840

2011 2,32,620 2,140 2,30,480 39,970 450 39,840

2012 2,29,060 2,190 2,26,870 39,920 410 39,510

2013 2,34,580 2,240 2,32,340 40,030 410 39,620

2014 2,35,030 2,360 2,32,670 40,430 430 40,000

In the Computer-Aided Detection (CAD), three anatomical landmarks i.e. pectoral region, breast frontiers and nipple are extracted automatically [6]. The current work is mainly concentrated on emphasizing the accuracy of pectoral muscle segmentation. Because, mammographic parenchyma and pectoral muscle has similar texture characteristics, which can be a source for misdiagnosis and cause high FP rate of breast cancer. There is large variation in texture, size, position, intensity and shape of the pectoral muscle, due to patient positioning during mammograms acquisition, also, have similar characteristics between breast tissue and pectoral muscle [6]. Therefore, detection of pectoral muscle is an important and challenging task.

2. Related Studies

To overcome the problems and improve the segmentation of pectoral muscle, various approaches are proposed and can be found in the literature [7-9], [11], [13-14].

Sreedevi S et al. [7] proposed a method for noise removal and pectoral muscle segmentation of mammographic images, using non local mean filter based on Discrete Cosine Transform (DCT) and canny edge detection in combination with global threshold respectively. Also, connected component labeling approach was used for suppression of pixels connected outside breast region. The method was tested for 161 mammograms from MIAS database, with an accuracy of 90.06%. The new method was proposed by R. J. Ferrari et al. [8] for automatic pectoral muscle segmentation in mammographic images, using Gabor wavelets based on multiresolution approach. The pectoral muscle enhancement in this approach was based on the convolution of pectoral region with designed Gabor filters. The algorithm was tested and evaluated on 84 mammograms for edge detection of pectoral muscle from MIAS database. Where, performance evaluation of the method was computed in percentage of FP rate 0.58% and FN rate 5.77% respectively. In this method segmentation of pectoral muscle is difficult, when pectoral muscle is hidden in the dense tissue. Since, detection and removal of pectoral muscle is an important aspect in mammograms, helps in image registration for analysis of the abnormalities from the breast. The combination of bit depth reduction and wavelet

analysis approach for pectoral muscle detection was proposed by Mario Mustra et al. [9]. The algorithm was tested for 40 mammograms with an accuracy of 85%. This method is suitable only, when the contrast of pectoral muscle is higher than that of surrounding tissue.

Wang et al. [11] presented a method for edge detection of pectoral muscle based on a active contour model and an discrete time Markov chain (DTMC). This method can overcome the disadvantages such as double layer muscles and line edges. But, the method is not suitable for segmentation of small pectoral muscle. David Raba et al. [14] suggested an algorithm to segment pectoral muscle, based on combination of region growing and morphological operation. If the contrast across breast tissue and pectoral muscle are indistinct, there may be over-segmentation of pectoral muscle with this approach. Sze Man Kwok et al. [13] proposed a method to detect the pectoral muscle based upon Hough transform and cliff detection. In this method, straight line concept was used first, to detect the edge of pectoral muscle roughly, and then polished to a curve. The limitation with this method, difficult to find correct muscle boundaries in some cases, due to complicated texture present in pectoral muscle.

Considering above facts, in this paper, we have proposed a simple intensity based detection algorithm for automatic breast boundary detection. The algorithm starts by convolution of mammograms with proposed filter enhancing the Region of Interest (ROI) having pectoral muscle. Finally, the boundary of pectoral muscle based on the seed point is obtained after enhancement processes, along with breast boundaries.

The paper is organized as follow. Introduction and current literature studies is presented in Section 1 and 2. Section 3 describes the concept of linear enhancement filter for enhancement and identification of pectoral region in mammograms. Experimental results and discussion based on the proposed method are illustrated in Section 4. Section 5 presents conclusion drawn based on the above experimentation.

3. Methodology

The distinct variation in intensity between pectoral region and breast tissue, approximately triangular structure, and gently tapers from top to bottom are some of the noteworthy anatomical features of the pectoral muscle [10]. An enhancement filter and frontier evolution approach based on the above listed features is proposed and considered in this paper to detect the boundaries of pectoral muscle automatically. The radiopaque artifacts i.e. (labels, wedges, etc.) can bias the detection of pectoral muscle. Hence, removal of artifact from the mammogram is important and achieved using pre-processing. To locate the pectoral muscle on the top left corner of the image, all right sided mammograms were flipped based on mean computation.

3.1. Pre-processing

The pre-processing is followed by artifact suppression i.e. (labels, wedges) from mammographic images using following steps:

Step 1: To reduce the noise present in the mammographic images, median filtering approach is used, as this technique, preserves the contrast and does not shift the boundaries of the image.

Step 2: Convert the filtered mammogram to binary using threshold technique, providing number of objects (artifacts and breast region) as an output.

Step 3: The breast region is obtained by computing the size of each object from step 2, considering only large object and removing small objects from the image.

Step 4: The pixel value 1 (white) is replaced by corresponding pixel value of original image in the binary mask generated in step 3, to extract the original pixel values of the mammograms.

3.2. Intensity Based Enhancement Filter

In the mammograms, pectoral muscles have high gray level intensities and approximate direction. Based on the previous comprehension, an enhancement mask is frequently used to enhance the pectoral muscle during processing of mammograms. In this work to increase the intensity of gray level of the pectoral muscle a linear enhancement mask with some coefficients is design and given by Eq (1):

Ie = y(i, j) = Fp q YZ1 ((i+ p, j - q) -1(i+ p, j + q)) + FI (i, j) (1)

p=0 q=1

where, Ie is the enhanced mammographie image, I(i, j) is the pixel intensity of an image at point (i, j), N are the number of image columns, The weighted differentiation effect for number of pixel pairs along vertical direction is denoted as M, and Fp, q and Fc are filter coefficients. The mammogram is convoluted with linear enhancement filter as shown in the Fig. 1, to implement the above equation. As pectoral muscle is gently tapers from top to bottom, the filter shown in Fig. 1(a) is rotated by 90 degree as shown in the Fig. 1(b), so as to emphasize the vertical edges of the pectoral region.

1 F c -1

1 0 -1

(a) (b)

Fig. 1 (a) Linear enhancement filter, (b) Linear enhancement filter rotated by 90 degree clockwise.

As, pectoral muscle have higher pixel intensity values compared to breast tissue with approximately triangular shape, which can be clearly depicted from Fig. 2(a).

Fig.2 (a) Pre-process Original mammogram case mdb055, (b) Enhanced mammographic image using linear enhancement filter, (c) pectoral boundary points obtained based on intensity variation using threshold (black dots).

The coefficient Fp, q enhance the triangular region of pectoral muscle, along with high intensity pixels from the mammogram and retains the rest of the pixel values from the breast region. Excluding centre point value Fc, addition of the filter coefficients is zero.

The pixel values from the pectoral muscles and similar pixel intensities in the original mammograms are enhanced, whereas rest of the pixel value remain same is the result of filter, depicted in Fig. 2(b). The centre point coefficient Fc is chosen in the range, 0 < Fc < 2, so as to retain the pixel values that are less than pixel value present in the pectoral muscle belongs to breast tissue, while processing mammograms.

3.3. Boundary point selection of the pectoral muscle

Based on the intensity difference characteristic of pectoral region and breast tissue, Jawad Nagi et al. [15] found four points iteratively on the pectoral boundaries using seed region growing approach and straight line equation for segmentation of the pectoral muscle. However, it is not easy to segment the pectoral muscles accurately, having variation in shapes using straight line concept. In order to obtain the accurate boundaries, intensity based approach is proposed for selecting the boundary points of pectoral muscle on the mammograms. The pectoral boundary point search and selection is explained in steps below:

1. Finding the maximum gray value of the enhanced image Ie and use the same as threshold value T.

2. Define M rows and N columns for the enhanced mammogram, scan row-wise with specific pixel interval from top to bottom, so as to obtain the pectoral boundary points.

3. Select and scan entire columns of first row from M rows, and store the first pixel location value of the row and column from (M, N) having intensity value less than threshold value T.

4. Increment the value of row with specific pixel interval towards downside of the mammogram, and repeat step. 1 for entire image.

5. The output of the step. 4 provides the location points on the pectoral boundaries, as depicted in Fig. 2(c) on pre-process original mammogram with black dots.

6. The location points obtained in step. 4 and shown in step. 5 are connected to segment the pectoral muscle from mammographic images.

3.4. Performance Analysis

The two parameters FP and FN rate [12] are used to evaluate the accuracy of the proposed method. The pixels detected outside the ground truth pectoral are defined as false positive pixels. The pixel detected inside the ground truth pectoral but not in the detected area is defined as false negative pixels. The FP and FN pixel rate are computed as below:

False Positive Pixel rate =

False Negative Pixel rate =

A U a 1 u 1 1 1 CI

x 100%

x 100%

Where, Aa e set of detected pectoral region and Ab e set of ground truth or reference region.

4. Experimental Results and Discussion

The proposed algorithm for segmentation of pectoral muscle from mammograms is described in this section. The proposed algorithm was tested using digitized mammogram with 200 ^m spatial resolution form mini-MIAS dataset. The detail of the database is presented in the Table 2 [16]. To locate the pectoral on the left top side all the right oriented mammograms were flipped. We have processed 320 mammograms using linear enhancement filter, since two mammograms does not show pectoral muscle i.e. mdb098 and mdb137, as per the opinion of radiologist.

Table 2. Detail of the database.




Spatial resolution

Gray-level quantization



Database Size

Database Tested

MLO (Medio-Lateral Oblique) 50 ^m /pixel 8 bits

1024 x 1024 pixels

Joyce-Loebl microdensitometer SCANDIG-3

The key basis to validate the proposed method was ground-truth of pectoral muscle provided by radiologists and their opinions, along with the quantitative analysis based on False Positive (FP) and False Negative (FN) rate. The proposed method elaborated in Fig. 3, has correctly enhanced and detected pectoral region that has been identified by a two radiologists. The parameters of the enhancement filter were set to 1 and -1 (except center coefficient Fc) to

have the zero sum of coefficients. The parameter Fc provides the intensity contribution of the current pixels. Usually, higher the value of Fc leads to superior enhancement of the pectoral region.

Figs. 4(b-d), 5(c), 6(c), 4(a), 5(a, b, d) and 6(a, b, d) demonstrate the results for fatty, fatty-glandular and dense glandular respectively, for some cases of pectoral muscle detection of MLO mammograms having small size, varying shapes, low contrast, double layer and dense mammograms using proposed technique. To assess the quality of muscle detection algorithm, the edges were manually drawn on each mammogram by the author and verified by two radiologists separately. These manually drawn muscle boundaries on the MLO view of mammograms were used as reference ground truth. The consensus was reached for the changes exist after discussion. The results of pectoral muscle detection were categorized into three classes as below:

• Accurate: In case of accurate results, pectoral muscle detected was similar to that of ground-truth.

• Acceptable: In case of acceptable results, more than 60% of pectoral region detected was from the ground-truth, with limited difference in case of glandular tissue present at the lower part of the mammogram.

• Inaccurate: The rest of the results were inaccurate and not correctly detected.

Pre-processing j

Fig.3 Schematic representation of automatic pectoral muscle detection.

The classification results are illustrated in Table 3 for 320 mammograms form mini-MIAS data set, with 96.56% of acceptable results with proposed algorithm. Furthermore, quantitative performance evolution was carried out on randomly selected 84 mammograms. The average FP and FN rate were 1.56 and 6.93 respectively.

Table 3. Pectoral boundary detection result classification for mini-MIAS database.

Category Number of mammograms Percentage (%)

Accurate 279 87.19

Acceptable 30 9.37

Inaccurate 11 3.44

As is apparent from Figs. 5(c) and 6(c) proposed method detects (black line) small pectoral region accurately. The effectiveness of the proposed approach, can be seen from Fig. 4(c) the pectoral region connected with suspicious mass is accurately detected. Figs. 5(d) and 6(b) illustrate the results for double layer and low contrast pectoral respectively.

Fig.4 Pectoral (black line) using for different cases (a) mdb001 (b) mdb025 (d)

muscles detected proposed method of MIAS database mdb009 (c) mdb028.

Fig.5 Pectoral muscles detected (black line) using proposed method for different cases of MIAS database (a) mdb035 (b) mdb057 (c) mdb070 (d) mdb110.

Fig.6 Pectoral .... ....n.. l . №)"»"■"» . ............<•" ' muscles detected

(black line) using proposed method for different cases of MIAS database (a) mdb111 (b) mdb138 (c) mdb150 (d) mdb201.

The prior knowledge about the information of the intensity and shape of the pectoral region is essential for detection of pectoral muscle using proposed method. The pectoral regions appear as, triangular with relatively high intensity on the upper corner of the mammogram. The intensity based technique is used, based on above characteristic for detection of the pectoral region. The simple filter mask of 3x2 has been proposed with few coefficients to enhance the pectoral region. The significance of the designed filter is, it considers the direction along with transition gray level intensity changes across the pectoral region compare to conventional enhancement filter. After pectoral enhancement, row and column wise search approach has been applied on the candidates to obtain the pectoral boundary points, based on threshold technique (black dots). The boundary points are finally connected to detect the pectoral muscle. However, the proposed approach fails in case if the entire mammogram has same gray level intensity, is the only limitation.

5. Conclusion

The pectoral muscle extraction is an important task, since it limits the search area for suspicious region in the mammogram and may bias in diagnosis procedures, due to same gray level intensity of pectoral muscle and abnormalities. Thus to overcome the above limitation, in this paper we have develop an algorithm for automatic pectoral boundary detection. In this approach, we have proposed a filter mask based on the characteristics of pectoral muscle for effective enhancement of the pectoral region. Then boundary points were obtained based on threshold technique. Finally, the boundary points obtained are connected to determine the boundary of pectoral muscle. The proposed algorithm has been tested and shows accuracy of 96.56% from radiologist experts over 320 mammograms from MAIS database.

The pectoral boundaries obtained using existing method is squiggly or irregular. Hence, in future work our focus is to develop an approach to refine the pectoral boundaries and test the algorithm on DDSM database.


The authors would like to thank Dr. Sushil Kachewar, Associate Professor, Dr. Sham Assistant Professor and their team, Department of Radiodiagnosis and Imaging of Rural Medical College, Pravara Institute of Medical Science (PIMS), Loni (Deemed University), for providing their timely consultation.



2. American cancer society, cancer facts and fgures, American Cancer Society,Atlanta, Ga, 2005-2015.

3. Karthikeyan Ganesan, URajendra Acharya, Chua Kuang Chua, Lim Choo Min, KThomas Abraham, and K Ng, Computer-aided breast cancer detection using mammograms: A review, Biomedical Engineering, IEEE Reviews in, 6:77-98, 2013.

4. Jinshan Tang, Rangaraj M Rangayyan, Jun Xu, Issam El Naqa, and Yongyi Yang, Computer-aided detection and diagnosis of breast cancer with mammography: recent advances. Information Technology in Biomedicine, IEEE Transactions on, 13(2):236-251, 2009.

5. Karthikeyan Ganesan, U Rajendra Acharya, Kuang Chua Chua, Lim Choo Min, and K Thomas Abraham. Pectoral muscle segmentation: a review. Computer methods and programs in biomedicine, 110(1):48-57, 2013.

6. Chunxiao Chen, Gao Liu, Jing Wang, and Gail Sudlow, Shape-based automatic detection of pectoral muscle boundary in mammograms. Journal of medical and biological engineering, 35(3):315-322, 2015.

7. S Sreedevi and Elizabeth Sherly. A novel approach for removal of pectoral muscles in digital mammogram. Procedia Computer Science, 46:17241731, 2015.

8. Ricardo J Ferrari, Rangaraj M Rangayyan, JE Leo Desautels, RA Borges, and Annie France Frere, Automatic identi_cation of the pectoral muscle in mammograms. Medical Imaging, IEEE Transactions on, 23(2):232-245, 2004.

9. Mario Mustra, Jelena Bozek, and Mislav Grgic, Breast border extraction and pectoral muscle detection using wavelet decomposition. In EUROCON 2009, EUROCON'09, IEEE, pages 1426-1433, 2009.

10. Yanfeng Li, Houjin Chen, Yongyi Yang, and Na Yang, Pectoral muscle segmentation in mammograms based on homogenous texture and intensity deviation. Pattern Recognition, 46(3):681-691, 2013.

11. Lei Wang, Miao-liang Zhu, Li-ping Deng, and Xin Yuan. Automatic pectoral muscle boundary detection in mammograms based on markov chain and active contour model. Journal of Zhejiang University SCIENCE C, 11(2):111-118, 2010.

12. Fei Ma, Mariusz Bajger, John P Slavotinek, and Murk J Bottema. Two graph theory based methods for identifying the pectoral muscle in mammograms. Pattern Recognition, 40(9):2592-2602, 2007.

13. Sze Man Kwok, Ramachandran Chandrasekhar, Yianni Attikiouzel, and Mary T Rickard. Automatic pectoral muscle segmentation on mediolateral oblique view mammograms. Medical Imaging, IEEE Transactions on, 23(9):1129-1140, 2004.

14. David Raba, Arnau Oliver, Joan Mart, Marta Peracaula, and Joan Espunya. Breast segmentation with pectoral muscle suppression on digital mammograms. In Pattern Recognition and Image Analysis, pages 471-478. Springer, 2005.

15. Jawad Nagi, S Abdul Kareem, Farrukh Nagi, and Syed Khaleel Ahmed. Automated breast profile segmentation for roi detection using digital mammograms. In Biomedical Engineering and Sciences (IECBES), 2010 IEEE EMBS Conference on, pages 87-92, IEEE, 2010.

16. John Suckling, J Parker, D Dance, S Astley, I Hutt, C Boggis, I Ricketts, E Stamatakis, N Cerneaz, S Kok, et al. The mammographic image analysis society digital mammogram database. In ExerptaMedica. International Congress Series, volume 1069, pages 375-378, 1994.