Detection of Tumour Based on Breast Tissue Categorization

Background: Breast cancer originates in breast tissue, which is made up of glands for milk production
(lobules), and the ducts that connect lobules to the nipple. Breasts contain both dense tissue
(glandular tissue and connective tissue, together known as fibro-glandular tissue) and fatty tissue.
Fatty tissue appears dark on a mammogram, whereas fibro-glandular tissue appears as white. Despite
the benefits of Computer Aided Detection (CAD), false detection of breast tumour is still a challenging
issue with oncologist. A mammography is a non-invasive screening tool that uses low energy X-rays to
show the pathology structure of breast tissue. Interpreting mammogram visually is a time consuming
process and requires a great deal of skill and experience. Earlier Computer Aided Techniques
emphasis detection of tumour in breast tissues rather than categorization of breast into Breast Imaging
Report and Data System (BI-RADS) which is the medically understandable method of reporting.
Aim: The work centred on developing a CAD system which is capable of not only detecting but also
categorizing breast tissue in line with BI-RADS scale.
Methodology: The acquired images were pre-processed to remove unwanted contents. Two stage
medical procedural approach was designed to categorize the tissue in breast images into low dense
(fatty) and high dense. Tumours in the low dense breasts were segmented, and then classified as
normal, benign and malignant. The developed system was evaluated using sensitivity, specificity,
false positive reduction, false negative reduction and overall performance.
Results: The developed CAD achieved 90.65% sensitivity, 73.59% specificity, 0.02 positive
reduction, 0.04 false negative reduction and 85.71% overall performance.
Conclusion: The false positive reduction result obtained shows that false detection has been
minimized as a result of categorization procedure of the breast tissue in mammograms. This article
has reported breast tumour detection from breast tissue categorisation using Medical procedural
approach. The developed system assisted in identification of suspicious mammograms and
identification of dense and fatty breasts. The classification of the segmented mammogram into
normal, benign and malignant achieved a better false positive reduction (0.02) and false negative
reduction (0.04) and thus provided an improved method for detection and classification of breast
tumour in terms of overall performance.