World Applied Sciences Journal 31 (6): 1056-1064, 2014
ISSN 1818-4952
© IDOSI Publications, 2014
DOI: 10.5829/idosi.wasj.2014.31.06.203
Corresponding Author: S. Pitchumani Angayarkanni, Department of Computer Science,
Lady Doak College, Madurai, Tamil Nadu, India.
1056
Mathematical Morphological Approach for Mammogram
Image Segmentation and Classification
S. Pitchumani Angayarkanni and Nadira Banu Kamal
1 2
Department of Computer Science, Lady Doak College, Madurai, Tamil Nadu, India
1
Department of M.C.A. TBAK College, Kilakarai, Ramnad,Tamil Nadu, India
2
Abstract: This paper presents the mathematical morphological and rough set based approach in detection and
classification of cancerous masses in MRI mammogram images. The main objective behind this approach is to
build a CAD system with good accuracy and computational speed in detection of cancerous masses compared
to the existing system. The ROI(Region of Interest) is segmented using Graph cut method.and the fourteen
features including morphological,shape and novel features are calculated for this region. Best rules used for
classification are generated using ID3 algorithm. Automatic classification based on the rules generated are
determined using Artificial bee colony based Multi Layered Perceptron model.The sensitivity, the specificity,
positive prediction value and negative prediction value of the proposed algorithm accounts to 98.79%, 98.8%,
92% and 96.6% which rates very high when compared to the existing algorithms. The area under the ROC curve
is 0.89. A GUI based tool was developed for the proposed methodology. An android application using simulator
was developed to make the doctor and patient to view the image with appropriate information like Patient Name,
age,Size of tumor, Nature of tumor and type of treatment.
Key words: Fuzzification Graph cut ID3 and Artificial Bee colony technique
INTRODUCTION problem if left unnoticed. Benign tumors are composed of
The Population Based Cancer Registry completely removed and are unlikely to recur. In MRI
evidently shows from the various statistics, that the mammogram images after the appropriate segmentation
incidence of breast cancer is rapidly rising, amounting of the tumor, classification of tumor into malignant,
to a significant percentage of all cancers in women. benign and normal is difficult task due to complexity
Breast cancer is the commonest cancer in urban areas and variation in tumor tissue characteristics like its shape,
in India and accounts for about 25% to 33% of all cancers size, grey level intensities and location. Feature extraction
in women. Over 50% breast cancer patients in India is an important aspect for pattern recognition problem.
present in stages 3 and 4, which will definitely impact the A Hybrid rough set based mathematical approach for
survival [1]. The survival rate can be increased only automatic detection and classification of cancerous
through the early diagnosis. Image processing technique masses in mammogram images is proposed in this
together with data mining is used for extraction and paper.
analysis of the ROI. Tumor can be classified into three
category normal, benign and malignant. A normal tumor MATERIALS AND METHODS
is a mass of tissue which exists at the expense of healthy
tissue. Malignant tumor has no distinct border. They tend The data set used for research were taken from
to grow rapidly increasing the pressure within the breast Mammogram Image Analysis Society (MIAS) [2]. The
cells and can spread beyond the point from they originate. database contains 320 images out of which 206 are normal
Grows faster than benign and cause serious health images, 63 benign and 51 malignant cases.
harmless cells, have clearly defined borders, can be