![](bg6.png)
sults using an adaptive bandwidth adjustment scheme
as well as follow-up stages for further improvement
of the initial clustering results. Due to the ability of
the adaptive mean shift in terms of working with non-
convex clusters as well as its noise smoothing behav-
ior, we are able to obtain good results after the final
segmentation. Moreover, the method was validated
on real mammography database and comparing to the
similar statistical approaches, it shows improvement
in sensitivity of mass detection and better false posi-
tives rate. One of the main objectives of this work is
to provide radiologists with a computer-aided detec-
tion system aimed at studying the risk of developing
breast cancer.
For further improvement of the proposed ap-
proach, we can consider the following works:
- Applying the method to a different data set.
- Applying parameter optimization methods.
- Using other classifiers such as SVM.
- Refinement of the segmentation method based on
shape attributes to capture all lesion extension.
Figure 7: Final results. The final segmentation after the
mode fusion is indicated by solid black contour where the
clusters are shown by different colors. (a) and (c) are
the results of the final segmentation; (b) and (d) are their
corresponding detected areas in the original breast images
(zoomed in for better visualization).
REFERENCES
Berman, C. G. (2007). Recent advances in breast-specific
imaging. Cancer control, 14(4):338–349.
Cheng, H. D., Shi, X. J., Min, R., Hu, L. M., Cai, X. P., and
Du, H. N. (2006). Approaches for automated detection
and classification of masses in mammograms. Pattern
Recognition, 39(4):646–668.
Cleveland, W. and Devlin, S. (1988). weighted regression:
An approach to regression analysis by local fitting.
Jour. of the American Stat.l Assoc., 83:596–610.
Comaniciu, D. and Meer, P. (2002). Mean-shift: A robust
approach toward feature space analysis. IEEE Trans-
action on Pattern Analysis and Machine Intelligence,
24(5):603–619.
Georgescu, B., Shimshoni, I., and Meer, P. (2003). Mean
shift based clustering in high dimensions: A texture
classification example. In IEEE International Confer-
ence on Computer Vision, pages 456–463. IEEE.
Horner, M. J., Ries, L., Krapcho, M., Neyman, N., Aminou,
R., Howlader, N., Altekruse, S. F., Feuer, E., Huang,
L., Mariotto, A., Miller, B., Lewis, D., Eisner, M.,
Stinchcomb, D., and Edwards, B. (2009). Seer can-
cer seer cancer statistics review, 1975-2006. National
Cancer Institute, Bethesda, MD, seer.cancer.gov.
Jimenez-Alaniz, J. R., Medina-Banuelos, V., and Yanez-
Suarez, O. (2006). Data-driven brain mri segmenta-
tion supported on edge confidence and a priori tissue
information. IEEE Transaction on Medical Imaging,
25(1):74–83.
Lee, J. (2007). Qualitya radiology imperative: Interpreta-
tion accuracy and pertinence. Journal of the American
College of Radiology, 4:162–165.
Mayer, A. and Greenspan, H. (2009). An adaptive mean-
shift framework for mri brain segmentation. IEEE
Transaction on Medical Imaging, 28(8):1238–1250.
Morton, M., Whaley, D., Brandt, K., and Amrami, K.
(2006). Screening mammograms: Interpretation with
computer-aided detectionprospective evaluation. Ra-
diology, 239(2):375–383.
Ojalaa, T., Nappia, J., and Nevalainena, O. (2001). Accu-
rate segmentation of the breast region from digitized
mammograms. Computerized medical imaging and
graphics, 25:47–59.
Sahba, F. and Venetsanopoulos, V. (2009). A new fuzzy
approach to mammographic breast mass segmenta-
tion. In International Conference on Image Pro-
cessing, Computer Vision, and Pattern Recognition
(IPCV2009), pages 852–858. Las Vegas.
Tomasi, C. and Manduchi, R. (1998). Bilateral filtering for
gray and color images. In International Conference
on Computer Vision, pages 839–846.
BREAST MASS DETECTION USING BILATERAL FILTER AND MEAN SHIFT BASED CLUSTERING
93