Breast Cancer Cell Detection Using Digital Image Processing

DOI : 10.17577/IJERTV1IS9046

Download Full-Text PDF Cite this Publication

Text Only Version

Breast Cancer Cell Detection Using Digital Image Processing

Prasanna G. Shete, Dr. Gajanan K. Kharate And Sanket C. Rege

Abstract The paper discusses an image processing method that aims at making more accurate estimates of the population of cancerous cells in the blood sample of a patient suffering from breast cancer. The method aims at providing a reliable, repeatable, and fast alternative that could replace the traditional method of manual examination and subsequent estimation. A marker, known as the Estrogen Receptor (ER), is used to identify cancer cells in the tissue sample. The methods involved in image processing are HSV color conversion from RGB image, Hue, saturation and value based object- background separation, morphological operations such as dilation and closing, and area based filtering for preliminary preparation of image for detailed analysis. Next, a modified watershed algorithm designed for eliminating errors arising due to over-segmentation in traditional watershed algorithm is proposed to provide comparatively more accurate results. Further, intensity based thresholding is performed for identifying and categorizing the cancerous cells into levels of severity of damage done to cells due to cancer. The proposed modified watershed algorithm is compared with the original watershed algorithm and an accuracy of almost 96% was observed and verified.

KeywordsCancer cells, Estrogen receptor, HSV model based object-background separation, intensity based cancer cell counting, modified watershed algorithm.



    reast cancer cells have receptors on their surface, in their cytoplasm and in the nucleus. Pathologists use external chemical hormones that bind to these receptors [1], and cause visible changes in the cell. The Estrogen receptor (ER) is one of the most reliable receptor which is actively used in cancer cell analysis and cell population estimation. Traditionally, the analysis is done by manually viewing the sample under microscope. As an alternate to this process, computer aided diagnosis involving image processing can be used to provide a reliable result with minimized errors [2], [3], [4]. The advantages of employing automation for pathological analysis over manual evaluation of cancer cell population are : Decreased time for analysis of pathological samples allowing

    Manuscript received 2nd November, 2012.

    1. Prasanna G. Shete is the Head of Department of Electronics and Telecommunication Engineering, PVGs COET, University of Pune, Pune, India.


    2. Dr. Gajanan K. Kharate is the Principal of Matoshri COER, Nasik, India.


    3. Sanket C. Rege is with the Department of Electronics and Telecommunication Engineering, PVGs COET, University of Pune, Pune, India.


    pathologists to avoid routine scanning and focus on other more complex issues, reduced number of errors the algorithms can be made highly accurate and can avoid false positives or misses giving a highly reliable result, faster documentation of results and higher repeatability due to the fact that medical images in digital format can be stored and reused for later analysis, and finally, minimization of costs as the entire process is automated, the cost per analysis is reduced.



    The method used currently involves using antigen- antibody reactions due to which cell staining takes place. The cancerous cells are stained in a dark brownish color (Fig. 1) while non-cancerous cells develop a bluish shade.

    Fig. 1: Progesterone receptor: stained tissue sample showing cancer positive/P (stained brown) and cancer negative/N (stained blue) cells.

    1. Receptors and markers

      A marker is a predictive indicator that helps to evaluate the response of cancer cells to a particular treatment. Of the many receptors identified [5]-[15], the estrogen receptor (ER) is of prime importance. Other important receptors frequently used in pathological analysis include progesterone receptor (PR) and HER2 [16] (Table I).

      TABLE I

      Various markers and their value in predicting breast cancer [17].

      Established and used in routine clinical analysis Potentially useful for clinical use; require


      Estrogen Receptor (ER) Epidermal Growth Factor Receptor (EGFR or HER1)

      Research interest, less likely to be used clinically


      Progesterone Receptor (PgR) Ki-67 (MIB-1) Cyclin E, Cyclin D1, p21, p27

      Human Epidermal Growth Factor Receptor (HER2)

      Topoisomerase II alpha Bcl2, bax, bcl-x, survivin

    2. Evaluation methods

    One of the most popular evaluation methods is the Immuno-histo-chemical or IHC method [18]. This method has a lot of advantages such as its ready availability, relative lower costs per analysis and simple methods for preservation of stained samples [19].

    The IHC method involves visual examination of cell membrane under a microscope. In HER2, evaluation further involves classification of tissue sample into categories of {0, 1+, 2+, and 3+} [16] depending on the severity of cell damage. Fluorescence-In-Situ-Hybridization (FISH) is another method for analyzing breast cancer tissue for HER2 receptor. Although it is more accurate than IHC method, it tends to be more expensive. Therefore we focus our research on the IHC method only.


    A number of approaches have been suggested by various authors for developing effective algorithms. In [20], the authors proposed a marker controlled watershed algorithm to avoid the over-segmentation errors while applying the classic watershed algorithm.

    Also, in [21], the authors made use of the marker controlled watershed algorithm which gave a fairly high overall percent correct agreement of 80%. However, the process of extraction of these binary markers is very difficult as too many of these markers cause over-segmentation errors while too few of them cause different objects to merge [22].

    Another technique for marker extraction was proposed by

    [23] based on "constrained region labeling" but again this is a complicated process.

    In [24], the authors proposed edge detection as well as intensity based extraction of objects of interest from background (also known as region growing). However, the edge detection method using techniques like sobel and canny are sensitive to noise while intensity based algorithms are computationally time consuming as each pixel's intensity is scanned in the image. In our algorithm reducing the errors is a higher priority rather than reducing the computation time and therefore we have adopted the intensity based region growing process.


    The implementation of image processing on an image of stained tissue sample involves the steps as shown in Fig. 2.

    Obtain image of stained tissue sample

    Convert acquired RGB image to HSV

    Separate objects (cells) from background using color thresholding

    Perform morphological operations

    Separate overlapping or touching cells using watershed algorithm

    Eliminate over-segmentation errors using proposed modification in watershed algorithm

    Classify cells on the basis of the intensities

    Fig. 2: Flowchart showing the steps involved in image processing implementation.

    1. Obtaining stained tissue sample image

      The pathologists acquire a tissue sample rom a patient under test and treat this tissue sample with hormones to produce antigen-antibody reactions causing cells to stain and undergo a change in color. The glass slide is then placed under a microscope capable of magnification up to 80X. A camera mounted on the eyepiece of the microscope can be used to capture the magnified image.

      The images used for our research were of ER stained samples in PNG format with an RGB color model and acquired using the process described above.

      MATLAB supports the PNG format and the image can be brought into the workspace by simply using the imread (filename.png) command. Fig. 3 (a) shows one of the images loaded into the MATLAB workspace.

      Vol. 1 Issue 9, November- 2012

      (a) (b)

      (c) (d)

      Fig. 3: Images showing original RGB image (a), corresponding HSV image (b), after color thresholding (c) and after morphological operations (d).

    2. Converting acquired RGB image into HSV

      The HSV color space (Hue, Saturation, and Value) is used because it provides much more details than the RGB color space does.

      The hue corresponds to all the colors: red, yellow, green, cyan, blue and magenta. With change in the magnitude of

      saturation, the variation from unsaturated (shades of gray) to fully saturated (absence of white component) occurs. The value component corresponds to various levels of brightness. Thus the HSV color model provides additional information regarding color, shade and brightness as compared to RGB color model.

      As can be seen in Fig. 3 (b), only the brown cancer positive cells have been highlighted by green, cell edges in blue and all cancer negative cells as well as the background in pink. The rgb2hsv (variable) command returns the HSV image to the user.

    3. Separating cells from background

      Now, the primary task is to extract only those regions that need to be considered for analysis, namely those regions where the cancerous cells are present. On observing the HSV image, (Fig. 3 (b)), we see that the green shaded regions correspond to the cells which need to be evaluated for counting.

      We separate the HSV image into its three channels namely hue, saturation and value and pixels that do not satisfy the

      green shaded regions are removed. The cell boundaries, as observed in Fig. 3(b), are represented in blue color. Thus by eliminating (by thresholding also known as region-growing based on intensities), [1] pink and blue regions and retaining only the green ones, separation of touching cells becomes an easier task in further analysis. This is an advantage of the HSV color scheme over the RGB color scheme [25], [26]. This process of classifying pixels to belong to a cell or background

      Vol. 1 Issue 9, November- 2012

      1. (b)

        Fig. 4: Images showing segmented image with over-segmentation errors (a), and rectified over-segmentation errors as well as classification into high intensity (red) and low intensity (blue) cells (b).

        generates a binary image with pixels satisfying the color range marked by logical ones (white) and the rest marked as logical zeroes (black). This returned image (Fig. 3 (c)) contains irregular object edges as well as holes which need to be removed using further morphological operations.

    4. Perform morphological operations

    The thresholding process returns the regions of cells but there is a problem of rough cell edges and holes in the bodies of the cells. During further analysis of the sample, these problems cause inaccurate results. Thus it is necessary to use the following morphological operations [27]: (1) Dilation, causing the surfaces to develop smoother edges and (2) Closing, causing holes in the surface of the cells to be filled in. Fig. 3 (d) shows the morphologically modified version of the original image.

    F. Eliminating over-segmentation errors

    As seen in Fig. 4 (a), there are a few errors due to over- segmentation. These errors arise when a single cell is divided into multiple regions due to intensity variations within the cell. Thus a single cell might be counted more than once due to these multiple segments.

    A simple algorithm was designed by us that calculated the distances between any two centroids of segments (Fig. 5). A minimum distance threshold value was calculated and all calculated distances were compared to this threshold (Equation (1)). The distances that were below this threshold value were identified as erroneous segments and were eliminated. The new centroid was calculated by averaging the coordinates of the error centroids.

    In MATLAB, operations such as imdilate (variable) and

    d x 2 x

    2 y 2 y


    i n


    imclose (variable) are used for dilation and closing respectively.

    E. Separating touching or overlapping cells

    The tissue sample will always contain a few cells that either overlap or just touch each others boundaries. Such cells may be evaluated to be a single cell; giving inaccurate results. A solution to this problem is to use the watershed algorithm [28]-[38] which, in simpler words, searches for regions with sudden changes in intensities and identify these as boundaries. The word watershed is a geographical term often used to refer to the region between two highlands where water collects during monsoon.

    Mathematically, the watershed algorithm is implemented by applying the distance transform [27]. The distance transform involves calculation of the separation of points in the image. In MATLAB, the bwdist (variable) function calculates the distance between each pixel whose value is zero and the nearest nonzero pixel for binary images.



    i i 1 i 1 i 1

    Where i represents the current segment under consideration and n represents the total number of segments.

    Fig. 5: Representative image showing the concept of elimination of over- segmentation errors

    Using (1), the distances were calculated in a number of iterations till all segments have been considered. Fig. 5 expounds, pictorially, the idea behind this algorithm.


    Summary of results of modified watershed algorithm obtained for 7 Estrogen receptor images as compared to the classic watershed algorithm.

    Image ID number

    Total number Percentage accuracy

    of cells of classic watershed

    Percentage error for

    classic watershed

    Percentage accuracy of

    modified watershed

    Percentage error for

    modified watershed















































    Total/Average %






    G. Categorizing cells on the basis of their intensities

    Pathologists often require the cells to be classified into categories on the basis of theintensities. This can be done by obtaining the RGB information at the center and around it for each cell followed by thresholding, Fig. 4 (b). In MATLAB, the regionprops (variable) command is used to obtain coordinates of recognized segments. Then the impixel (variable, coordinates) is used to obtain RGB information at that point from the original image. The RGB information is then compared to fixed thresholds corresponding to different intensity levels.


    The count obtained by the modified watershed algorithm was compared to the count obtained by manual analysis by the pathologist. Table II shows a summary of the results obtained when the modified watershed algorithm was implemented on seven of the images that were provided as well as the results of the original watershed algorithm implemented on the same set of images. An average accuracy of 96.44% can be obtained in this way. The table also shows the results for the implementation of the classic watershed algorithm. The percentage accuracy in this case is less than 50% indicating a high number of errors due to over-segmentation. There is a possibility where our technique may fail. If a cell of smaller size is on the periphery of a cell containing the over- segmentation error or in the case of an elongated cell (Fig. 6),

    Fig. 6: Incorrectly interpreted distances

    the algorithm might wrongly interpret the first situation as an error due to small inter-centroid distance and the second situation having no error due to larger inter-centroid distance, both instances causing an unacceptable error in count.

    A remedy to this is to involve the user in a review process that allows him/her to either confirm or reject the result produced for each instance.

    The results are also affected by the quality of the image being analyzed. Images, in which tissue samples are folded, blurred or non-uniformly illuminated, produce erratic results. The solution to this is to have a proper image normalization or de-blurring process prior to image analysis, as the case may be.

    Fig. 7 shows the graphical summary of implementation of the modified watershed algorithm on Image ID number 5.


    As prescribing the treatment to a cancer patient is a matter of great responsibility, the results presented to the pathologists must be accurate and reliable. An error due to a miss or a false positive result may be catastrophic if the wrong medication is administered to the patient.

    It will still be a while before a pathologist will completely trust an automated algorithm to make decisions on his/her behalf. There may be a possibility in the future when the algorithm might perform even better than a pathologist in making decisions on the basis of numerical data instead of the pathologists instinct.

    The algorithm is, and will always be, only a tool used by pathologists to help them in the tedious work of counting cell population and not a replacement of the pathologist. The time saved by pathologists can be better utilized to address even greater problems in the field of bio-medical sciences.

    Fig. 7: The breast cancer tissue showing the implementation of modified watershed algorithm as well as intensity based categorization. The cells marked in green are lower in intensity while the cells marked in red are of higher intensity.


    The threshold level for minimum distance used in the algorithm is manually chosen for now as this set value satisfies the images when factors such as magnification remain fairly constant. For the algorithm to work with different magnification levels, a new threshold value needs to be manually evaluated. This process of calculating an optimal distance can be automated for a more generalized implementation of the algorithm.


    The authors gratefully acknowledge the guidance of Dr. Aparna Joshi, MD pathologist, consultant histo-pathologist, KEM hospital, Pune, India, in the field of immuno-histo- chemical analysis of tissue samples and providing the images used in this paper.

    The authors gratefully acknowledge the guidance of Dr. Mayur Kulkarni, student MD radiologist, Sassoon hospital, Pune, India.


  1. Pornchai Phukpattaranont and Pleumjit Boonyaphiphat, Color Based Segmentation of Nuclear Stained Breast Cancer Cell Images, in ECTI Transactions on Electrical Engineering, Electronics, and Communications, vol. 5, no. 2, pp. 158, August 2007.

  2. J. P. Thiran and B. Macq, Morphological feature extraction for the classification of digital images of cancerous tissues, in IEEE Transactions on Biomedical Engineering., vol. 43, no. 10, pp. 1011- 1020, October 1996.

  3. B. Fang, W. Hsu, and M. L. Lee, On the accurate counting of tumor cells, in IEEE Transactions on Nanobioscience., vol. 2, no. 2, pp. 94- 103, June 2003.

  4. S. Petushi, C. Katsinis, C. Coward, F. Garcia, and A. Tozeren,

    Automated identification of microstructures on histology slides, in IEEE International Symposium on Biomedical Imaging : Macro to Nano, vol. 1, pp. 424-427, 2004.

  5. P. L. Fitzgibbons, D. L. Page, D. Weaver, A. D. Thor, D. C. Allred,

    G.M. Clark, S. G. Ruby, F. OMalley, J. F. Simpson, J. L. Connolly,

    D.F. Hayes, S. B. Edge, A. Lichter, and S. J. Schnitt, Prognostic factors in breast cancer: College of americal pathologists consensus statement 1999, in Archives of Pathology and Laboratory Medicine, vol. 124, pp. 966978, 2000.

  6. J. S. Ross, G. P. Linette, J. Stec, E. Clark, M. Ayers, N. Leschly, W. F. Symmans, G. N. Hortobagyi, and L. Pusztai, Breast cancer biomarkers and molecular medicine, in Expert Review of Molecular Diagnostics, vol. 3, pp. 573585, 2003.

  7. M. Colozza, E. Azambuja, F. Cardoso, C. Sotiriou, D. Larsimont, and

    M. J. Piccart, Proliferative markers as prognostic and predictive tools in early breast cancer: Where are we now, in Annals of Oncology, vol. 16, pp. 17231739, 2005.

  8. S. P. Linke, T. M. Bremer, C. D. Herold, G. Sauter, and C. Diamond, A multi-marker model to predict outcome in tamoxifen-treated breast cancer patients, in Clinical Cancer Research, vol. 12, pp. 11751183, 2006.

  9. H. Yamashita, M. Nishio, T. Toyoma, H. Sugiura, Z. Zhang, S. Kobayashi, and H. Iwase, Coexistence of HER2 over-expression and p53 protein accumulation is a strong prognostic molecular marker in breast cancer, in Breast Cancer Research, vol. 6, pp. 2430, 2004.

  10. D. Coradini and M. G. Daidone, Bio-molecular prognostic factors in breast cancer, in Current Opinion Obstetrics Gynaecology, vol. 16, pp. 4955, 2004.

  11. S. Bose, S. Chandran, J. M. Mirocha, and N. Bose, The AKT pathway in human breast cancer: A tissue-array-based analysis, in Modern Pathology, vol. 19, pp. 238245, 2006.

  12. G. Perez-Tenorio and O. Stal, Activation of AKT/PKB in breast cancer predicts a worse outcome among endocrine treated patients, in British Journal of Cancer, vol. 86, pp. 540545, 2002.

  13. K. J. Schmitz, F. Otterbach, R. Callies, B. Levkau, M. Holscher, O. Hoffmann, F. Grabellus, R. Kimmig, K. W. Schmid, and H. A. Baha,

    Prognostic relevance of activated AKT kinase in node-negative breast cancer: A clinic-pathological study of 99 cases, in Modern Pathology, vol. 17, pp. 1521, 2004.

  14. J. N. Hutchinson, J. Jin, R. D. Cardiff, J. R.Woodgett, and W. J. Muller,

    Activation of AKT-1 (PKB-a) can accelerate ErbB-2 mediated mammary tumotigenesis but suppresses tumour invasion, in Cancer Research, vol. 64, pp. 31713178, 2004.

  15. S.Wullschleger, R. Loewith, and M. N. Hall, TOR signalling in growth and metabolism, in Cell, vol. 124, pp. 471484, 2006.

  16. Hela Masmoudi, Stephen M. Hewitt, Nicholas Petrick, Kyle J. Myers, and Marios A. Gavrielides, Automated Quantitative Assessment of HER-2/neu Immuno-histo-chemical Expression in Breast Cancer, in IEEE trasactions on medical imaging, vol. 28, no. 6, pp. 916, June 2009.

  17. R. A. Walker, Immuno-histo-chemical markers as predictive tools for breast cancer, in Journal of Clinical Pathology, pp. 689-696, originally published online November 23, 2007.

  18. Z. Ahmed, N. S. Azad, Y. Bhurgari, R. Ahmed, N. Kayani, S. Pervez, and S. Hasan, Significance of immune-histo-chemistry in accurate characterization of malignant tumours, in J. Ayub. Medical College, Abbottabad, vol. 18, pp. 3843, 2006.

  19. Z. Theodosiou, I. Kasampalidis, G. Livanos, M. Zervakis, I. Pitas, and

    K. Lyroudia, Automated analysis of fish and immuno-histo-chemistry images: A review, in Cytometry A, vol. 71, pp. 459450, 2007.

  20. H. S. Wu and J. Barba, An algorithm for noisy cell contour extraction via area merging, in Journal of Imaging Science and Technology, vol. 38, pp. 604-607, November 1994.

  21. Marios A. Gavrielides, Hela Masmoudi, Nicholas Petrick, Kyle J. Myers, and Stephen M. Hewitt, Automated Evaluation of HER-2/neu Immuno-histo-chemical Expression in Breast Cancer using Digital Microscopy, in Biomedical Imaging: From Nano to Macro, pp. 808- 811, May 2008.

  22. Hai Gao, Wan-Chi Siu, and Chao-Huan Hou, Improved Techniques for Automatic Image Segmentation, in IEEE Transactions on Circuits and Systems for Video Technology, vol. 11, no. 12, pp. 1273-1280, December 2001.

  23. P. Salembier and M. Pardas, Hierarchical morphological segmentation for image sequence coding, in IEEE Transactions on Image Processing, vol. 3, pp. 639651, September 1994.

  24. R. M. Haralik and L. G. Shapiro, Survey: Image segmentation techniques, in Computer Vision, Graphics and Image Processing, vol. 29, pp. 100-132, 1985.

  25. Cheng, H.D. and Y.Sun, A Hierarchical Approach to Color Image Segmentation Using Homogeneity, in IEEE Transactions on Image Processing, vol. 9, no. 12, pp. 2071-2082, 2000.

  26. Cheng, Y.Sun, Mean Shift, Mode Seeking, and Clustering, in IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 17, no. 8, pp. 790-799, 1995.

  27. J. M. Sharif, M. F. Miswan, M. A. Ngadi, Md Sah Hj Salam, Muhammad Mahadi bin Abdul Jamil, Red Blood Cell Segmentation Using Masking and Watershed Algorithm: A Preliminary Study, in 2012 International Conference on Biomedical Engineering (ICoBE), Penang, Malaysia, pp. 27-28, February 2012.

  28. N. Malpica, C. de Solrzano, I. J.Vaquero,A. Santos, I.Vallcorba, J. Garca-Sagredo, and F. del Pozo, Applying watershed algorithms to the segmentation of clustered nuclei, in Cytometry, vol. 28, pp. 289297, 1997.

  29. J. B. T. M. Roerdink and A. Meijster, The watershed transform: Definitions, algorithms and parallelization strategies, in Fundamental Informatics, vol. 41, pp. 187228, 2001.

  30. C. Ortiz de Solorzano, E. Garcia Rodriguez, A. Jones, D. Pinkel, J.W. Gray, D. Sudar, and S. J. Lockett, Segmentation of confocal microscope images of cell nuclei in thick tissue sections, in Journal of Microscopy, vol. 193, no. 3, pp. 212226, March 1999.

  31. H. Ancin, B. Roysam, T. E. Dufresne, M. M. Chestnut, G. M. Ridder, D.

    H. Szarowski, and J. N. Turner, Advances in automated 3-D image analyses of cell populations imaged by confocal microscopy, in Cytometry, vol. 25, no. 3, pp. 221234, November, 1996.

  32. M. K. Chawla, G. Lin, K. Olson, A. Vazdarjanova, S. N. Burke, B. L. McNaughton, P. F.Worley, J. F. Guzowski, B. Roysam, and C. A. Barnes, 3D-catFISH: a system for automated quantitative three- dimensional compartmental analysis of temporal gene transcription activity imaged by fluorescence in situ hybridization, in Journal of Neuroscience Methods, vol. 139, no. 1, pp. 1324, October, 2004.

  33. G. Lin, U. Adiga, K. Olson, J. F.Guzowski, C. A. Barnes, and B. Roysam, A hybrid 3D watershed algorithm incorporating gradient cues and object models for automatic segmentation of nuclei in confocal image stacks, in Cytometry A, vol. 56, no. 1, pp. 2336, November 2003.

  34. G. Lin, M. K. Chawla, K. Olson, C. A. Barnes, J. F. Guzowski, C. Bjornsson, W. Shain, and B. Roysam, A multi-model approach to simultaneous segmentation and classification of heterogeneous populations of cell nuclei in 3D confocal microscope images, in Cytometry A, vol. 71, no. 9, pp. 724736, September 2007.

  35. G. Lin, M. K. Chawla, K. Olson, J. F. Guzowski, C. A. Barnes, and B. Roysam, Hierarchical, model-based merging of multiple fragments for improved three-dimensional segmentation of nuclei, in Cytometry A, vol. 63, no. 1, pp. 2033, 2005.

  36. C. Wahlby, I. M. Sintorn, F. Erlandsson, G. Borgefors, and E. Bengtsson, Combining intensity, edge and shape information for 2D and 3D segmentation of cell nuclei in tissue sections, in Journal of Microscopy Oxford, vol. 215, pp. 6776, July 2004.

  37. R. W. Mackin, B. Roysam, T. J. Holmes, and J. N. Turner, Automated three-dimensional image analysis of thick and overlapped clusters in cytologic preparations. Application to cytologic smears, in Analytical and Quantitative Cytology and Histology, vol. 15, no. 6, pp. 405417, December 1993.

  38. R.W. Mackin, Jr., L.M. Newton, J. N. Turner, and B. Roysam,

Advances in high-speed, three-dimensional imaging and automated segmentation algorithms for thick and overlapped clusters in cytologic preparations. Application to cervical smears, in Analytical and Quantitative Cytology and Histology, vol. 20, no. 2, pp. 105121, April 1998.

Vol. 1 Issue 9, November- 2012

Prasanna G. Shete (M06) received the

B.E. degree in electronics engineering from the Walchand Institute of Technology, Solapur, Maharashtra, India, in 1991, and the M.Tech degree in digital systems from the Government College of Engineering, Pune, Maharashtra, India, in 2006. He is currently pursuing his Ph.D degree in biomedical image processing from the Sinhgad Research Centre, University of Pune, India.

He has been with the SSVPS College of Engineering, Dhulia, in the Electronics and Telecommunication Engineering

Department as a Lecturer from 1991 to 1994. Since 1994, he has been with PVGs College of Engineering and Technology, Pune, India as a Lecturer from 1994 to 2007, as an Assistant Professor from 2007 till date in the Electronics and Telecommunication Engineering Department and is currently also the Head of Department. He has published papers titled Intrusion Detection System (IDS) using Network Processor in Innovative Applications of Information Technology for the Developing World, December 2005 at Kathmandu, Nepal, and Intrusion Detection System (IDS) using Network Processor (IXP 2400), in the National Conference on Exploring the Latest Technological Trends, November 2005. His current research interest includes biomedical breast cancer marker evaluation using digital image processing under the guidance of Dr. Gajanan Kharate.

Mr. Shete is a Life Member of Indian Society for Technical Education (ISTE) since the year 2000.

Dr. Gajanan K. Kharate received the

B.E. degree in electronics engineering from the SSGMCE, Amarawati University, Shegaon, India, in 1987, the

M.E. degree in electronics engineering from Walchand College of Engineering, Shivaji University, Kolhapur, India in 1997 and the Ph.D degree in electronics and telecommunication engineering from University of Pune, Pune, India in 2007.

He has been with the K.K. Wagh Institute of Engineering Education and Research, Nashik, India as a Professor in

Electronics Engineering as well as the Head of Electronics Engineering Department from 1998 to 2002 and Head of IT Department from 2002 to 2008. Since 2008, he has been the Principal of Matoshri College of Engineering and Research Center, Nashik, India, serves as Dean, Faculty of Engineering, University of Pune, and an approved Ph.D Teacher for Electronics Engineering. He has published several papers such as Image Compression usingWavelet Packet Tree, in International Journal of Graphics, Vision and Image Processing, vol-5, issue-7, July 2005 and presented the papers titled Image Processing at the National Conference on Recent Advancements in Mechanical Engineering, Nashik, India, January 2004 and Multi Wavelet Image Compression at National level conference on Multimedia Technology and Applications, Coimbatore, India, July 2004. His current research interests include Digital Signal Processing, Image Processing and VLSI Design.

Dr. Kharate is a Life Member of Indian Society for Technical Education (ISTE), Fellow Member of Institution of Electronics and Telecommunication Engineers (IETE), Life Member of Computer Society of India and Life Member of IE (I).

Sanket C. Rege is currently pursuing the

B.E. degree in electronics and telecommunication engineering from the PVGs College of Engineering and Technology, Pune, India.

He is currently engaged in a research project on cancer cell (Estrogen) population estimation using digital image processing in collaboration with Dr. Aparna Joshi at the King Edward Memorial Hospital, Pune, India and under the guidance of Prasanna G. Shete. He is also working towards enhancing techniques involved in digital image analysis as well as in machine

learning algorithms.

Leave a Reply