Background: Breast cancer remains the leading cause of cancer deaths among women
globally. To reduce the breast cancer mortality, early detection, diagnosis and treatment is an important
requirement. Computer Aided Diagnosis (CADx) techniques with screening mammography
are widely used for this purpose. Both Medio Lateral Oblique (MLO) and Cranio Caudal (CC)
view mammograms are used in CAD system to improve the performance.
Objective: To propose a diagnosis method for accurate detection of breast cancer using fusion of
mammographic texture features.
Methods: In this study, fusion of Local Binary Patterns (LBP) or binary Gabor patterns (BGP) of
MLO and CC view images using Canonical Correlation Analysis (CCA) is proposed to improve
the diagnostic accuracy and to reduce the false positive rate. Two data bases, Digital Database for
Screening Mammography (DDSM) and INbreast are used to evaluate the performance of the proposed
Results: Compared to baseline system, the proposed the CCA based serial fusion improves the accuracy
by 6.425% and 6.395% for DDSM and INbrest databases respectively.
Conclusion: BGP is robust to rotations and gray-scale invariant and produces significant texture
features to characterize tumors in a better manner. As CCA transforms the features in a way that
the transformed features have maximum cross correlation and minimum auto correlation, the CCA
based serial fusion improves the system performance. The proposed system could help the radiologists
in early diagnosis and therefore reduce the mortality.