Breast cancer is an uncontrolled growth of breast cells. Cancer occurs as a result of mutations, or abnormal changes, in the genes responsible for regulating the growth of cells and keeping them healthy. The genes are in each cell’s nucleus, which acts as the “control room” of each cell. Normally, the cells in our bodies replace themselves through an orderly process of cell growth: healthy new cells take over as old ones die out. But over time, mutations can change cells by “turning on” certain genes and “turning off” others in a cell. Changed cell gains the ability to keep dividing without control, producing more cells just like it and forming a tumor (World Health Organization International Agency for Research on Cancer, June 2003; a b c World Health Organization, February 2006).

A tumor can be benign (not dangerous to health) or malignant (dangerous to health). Benign tumors are not considered cancerous: their cells are close to normal in appearance, they grow slowly, and they do not invade nearby tissues or spread to other parts of the body. Malignant tumors are cancerous. Less commonly, breast cancer can begin in the stromal tissues (G.L Takkab, 2003) which include fatty and fibrous connective tissues of the breast. It is widely believed that the breast cancer is caused by a genetic abnormality (http://www.merck.com/mmhe/sec23/ch266/ch266a.html). However, only 5-10% of cancers are due to an abnormality inherited from mother or father. About 90% of breast cancers are due to genetic abnormalities that happen as a result of the aging process and the “wear and tear” of life in general. One of the outstanding problems in the present day is to diagnose, define and classify the type of breast cancer using minimum cost. Several Artificial Intelligence (AI) techniques including neural networks and fuzzy logic are successfully applied to a wide variety of decision making problem in the area of medical diagnosis.

This paper is organized as follows: Section 2 describes Artificial Neural Network. Section 3 describes Wisconsin Breast Cancer Database. Section 4 describes related work to this research. Section 5 describes Algorithm part. Section 6 describes Simulation and results. The final section provides some conclusion relating to the performance of the algorithm when applied to the breast cancer.

During the recent times, because of their discriminative training ability and easy implementation, the Artificial Neural Networks (ANN) find extensive use in classification of the type of tumor in breast cancer problem. It turns out that the selection of number of nodes for an ANN is an important criterion in breast cancer analysis. However, a large network means more computational expenses, resulting in more hardware and time related cost. Therefore a compact and optimum design of neural network is needed towards real time detection of tumor type in breast cancer analysis.

An ANN can be defining as a highly connected array of elementary processors called neurons. A widely used model called the multi-layered perceptron (MLP) Figure 1. The MLP type ANN consists of one input layer, one or more hidden layers and one output layer. Each layer employs several neurons and each neuron in a layer is connected to the neurons in the adjacent layer with different weights. Signals flow into the input layer, pass through the hidden layers, and arrive at the output layer. With the exception of the input layer, each neuron receives signals from the neurons of the previous layer linearly weighted by the interconnect values between neurons. The neuron then produces its output signal by passing the summed signal through a sigmoid function.

The Wisconsin Breast Cancer dataset (Merz J, 1996) was obtained from a repository of a machine earning Database. The Wisconsin Prognostic Breast Cancer (WDBC) dataset contains 69 instances with 32 attributes each. There are no missing attribute in the dataset. This dataset can be use to predict if breast tumor is benign and malignant.

Several studies have been conducted based on this database. For example, Bennet and Mangasarian (Nett K.P, 1992) used linear programming techniques, obtaining a 99.6% classification rate on 487 cases (the reduced database available at the time) (Ravi Jain, 2003).

Have comparative study of fuzzy classification methods on Breast Cancer Data. There are 9 attributes which are primarily used to classify the breast cancer as either benign or malignant based on cell description gathered by microscopic examination. These are: 1. Clump Thickness 2. Uniformity of Cell Size 3. Uniformity of Cell Shape 4. Marginal Adhesion 5. Single Epithelial Cell Size 6. Bare Nuclei 7. Bland Chromatic 8. Normal Nucleoli and 9. Mitoses.

Breast cancer data has been divided into two categories: one dataset is used for training and other dataset is used for testing. Normalized Data is posted (http://www.scribd.com/doc/14939420/Refrencing-Data.html). Multilayer neural network is trained and tested using 569 Data Samples (http://www.scribd.com/doc/14939420/Refrencing-Data.html). In Training Data (http://www.scribd.com/doc/14939420/Refrencing-Data.html) the first column is ID number, columns 2-10 represent characteristics of cancer cell, and the 11th field denotes the type of cancer (2 denotes malignant and 4 denotes benign).

The type of Breast cancer is decided by all 9 attributes (http://www.scribd.com/doc/14939420/Refrencing-Data.html).

The artificial neural network has become a very popular alternative in prediction and classification tasks due to its associated memory characteristics and generalization capability (Shien-Ming Chon, 2004). Thirteen cytology of fine needle aspiration image (i.e. cellularity, background information, cohesiveness, significant stromal component, clump thickness, nuclear membrane, bare nuclei, normal nuclei, mitosis, nucleus stain, uniformity of cell, fragility and number of cells in cluster) are evaluated their possibility to be used as input data for artificial neural network in order to classify the breast precancerous cases into four stages, namely malignant, fibro adenoma, fibrocystic disease, and other benign diseases. An intelligent diagnostic system based on the hybrid multilayer perceptron (HMLP) network to determine the four stages of breast pre-cancerous, namely malignant, fibroadenoma, fibrocystic disease, and other benign diseases was proposed (Nor Ashidi MatIsa, 2007). An automated computerized classification method based on computer vision and artificial neural networks to estimate the likelihood that a mammographic lesion is malignant. The features selected to characterize the mass lesions as well as the classifier used to merge the features are crucial components of the overall classification method. We have shown that our computerized method is robust to case mix and digitization technique (Maryellen L. Giger, 2000). The three neural networks were trained and compared using 1300 data samples. The classification results are indicating that all the networks give good overall diagnostic performance. However, only Hybrid Multilayered Network that provides 100% accuracy, sensitivity and specificity (Mohd Yusoff Mashor, 1996). We have developed a computer vision system that can classify microcalcifications (Yuzheng C. Wu, 1995) objectively and consistently to aid radiologists in the diagnosis of breast cancer. A convolution neural network (CNN) was employed to classify benign and malignant microcalcifications in the radiographs of pathological specimen.

In this research MATLAB is used for implementation. The training data is stored in the file Train.txt which contains training patterns for patients. Firstly, feed forward neural network is trained with 2 to 10 columns as inputs and 11th column as output. Then the network is simulated with input and output. After that testing of network is done by taking Data from test.txt file (contains the testing data). The algorithm is as follows:

dex= load('train.txt');

p=x(:,2:10);

p=p';

t=x(:,11);

t=t';

network=newff(minmax(p),[9,6,1],

{'logsig','logsig','logsig'},'traingdm');

net.trainParam.lr=0.5; net.trainParam.mc=0.9;

network=init(network);

y=sim(network,p);

figure,plot(p,t,p,y,'o'),title('Before Training');

network.trainParam.epochs = 100;

network=train(network,p,t);

y=sim(network,p);

figure,plot(p,t,p,y,'o'),title('after Training');

display('Final weight vector and bias values : \n');

Weights=network.iw{1}; Bias=network.b{1};

Weights

Bias

y=load('test.txt'); p1=y(:,2:10);

p1=p1'; t1=y(:,11); t1=t1';y1=sim(network,p1);

figure,plot(p1,t1,p1,y1,'o'),title('after testing');

If the algorithm is executed for all attributes (inputs) in MATLAB, Training is given for all 9 parameters; it gives accurate results as shown in Figure 2 and Figure 3. It means that type of breast cancer is decided accurately. Bottom X axis represents Actual Output. Top X axis Represents Target Output. Left and Right Y Axis Represents Input.

If a particular attribute is removed then dotes (colored) coincides on the point 1 at the top horizontal than output is accurate and this means that particular attribute can be discarded. Sample Data in Table 2 shows the mathematical deviation in accuracy of output after removing an attribute. For complete data refer to reference (http://www.scribd.com/doc/14939420/Refrencing-Data.html).

In this paper, a case study of attributes is done for real time detection of tumor type in breast cancer analysis. Our study shows that uniformity of cell size, single epithelial cell size and Normal Nucleoli attributes are very important to decide tumor type. But clump thickness, uniformity of cell shape, bare nucleoli, Marginal Adhesion, bland chromatic and mitoses can be ignored. Effectively reducing the time and cost in analyzing tumor type. It can be seen that in previous researches, only the classification of cancer is being studied. In this research, the significant role of each attribute is discussed and studied using matlab. This can help in Hospitals.

The advantage of the proposed algorithm is that it is very simple to implement, neither the trial-and-error process nor prior knowledge about the parameters is required.