BRAIN AND LUNG CANCER DETECTION

OBJECTIVE:
To identify brain tumor and lung cancer by using computer images.

ABSTRACT: Segmentation of pulmonary X-ray computed tomography (CT) images is a precursor to most pulmonary image analysis applications. Digital Image Processing is currently a hot research area in medicine and it is believed that they will receive extensive application to biomedical systems in the next few years. In Digital Image Processing, neural networks are ideal in recognizing diseases using scans since there is no need to provide a specific algorithm on how to identify the disease. Computer aided diagnosis systems for detecting malignant texture in biological study have been investigated using several techniques. This is an approach in computer-aided diagnosis for early prediction of brain cancer using Texture features and neuro-classification logic and an algorithm to separate the lung tissue from a Chest CT to reduce the amount of data that needs to be analyzed.

INTRODUCTION
In recent years, the occurrence of brain tumors has been on the rise. Unfortunately, many of these tumors will be detected too late, after symptoms appear. It is much easier and safer to remove a small tumor than a large one. About 60 percent of glioblastomas start out as a lower-grade tumor. But small tumors become big tumors. Low-grade gliomas become high-grade gliomas. Once symptoms appear, it is generally too late to treat the tumor. Computer-assisted surgical planning and advanced image-guided technology have become Increasingly used in Neuro surgery. The availability of accurate anatomic three-dimensional (3D) models substantially improves spatial information concerning the relationships of critical structures (eg, functionally significant cortical areas, vascular structures) and disease [6]. In daily clinical practice, however, commercially available intraoperative navigational systems provide the surgeon with only two-dimensional (2D) cross sections of the intensity-value images and a 3D model of the skin. The main limiting factor in the routine use of 3D models to identify (segment) important structures is the amount of time and effort that a trained operator must spend on the Preparation of the data. A brain cancer is a disease in which cells grow uncontrollably in the brain. Brain tumors are of two main types : 1) Benign tumors 2) Malignant tumors Benign tumors are incapable of spreading beyond the brain itself. Benign tumors in the brain usually do not need to be treated and their growth is self limited. Sometimes they cause problems because of their location and surgery or radiation can be helpful.

Malignant tumors are typically called brain cancer. These tumors can spread outside of the brain. Malignant tumors of the brain will always develop into a problem is left untreated and an aggressive approach is almost always warranted. Brain malignancies can be divided into two categories : Primary brain cancer originated in the brain. Secondary or metastatic brain cancer spreads to the brain from another site in the body. Cancer occurs when cells in the body ( in this case brain cells ) divide without control or order. Normally, cells divide in a regulated manner. If cells keep dividing uncontrollably when new cells are not needed, a

mass of tissue forms, called a growth or tumor. The term cancer usually refers to malignant tumors, which can invade nearby tissues and can spread to other parts of the body.

Lung cancer is a disease that occurs because of uncontrolled cell growth in tissues of the lung. This growth may lead to metastasis, which is the invasion of adjacent tissue and infiltration beyond the lungs. Treatment and prognosis depend on the histological type of cancer, the stage (degree of spread), and the patient's performance status. Possible treatments include surgery, chemotherapy, and radiotherapy. Survival depends on stage, overall health, and other factors, but overall only 14% of people diagnosed with lung cancer survive five years after the diagnosis. Symptoms that may suggest lung cancer include: dyspnea (shortness of breath with activity), hemoptysis (coughing up blood), chronic coughing or change in regular coughing pattern, wheezing, chest pain or pain in the abdomen, cachexia (weight loss, fatigue, and loss of appetite), dysphonia (hoarse voice), clubbing of the fingernails(uncommon), dysphasia(difficulty swallowing), Pain in shoulder ,chest , arm, Bronchitis or pneumonia, Decline in Health and unexplained weight loss.

If the cancer grows in the airway, it may obstruct airflow, causing breathing difficulties. The obstruction can lead to accumulation of secretions behind the blockage, and predispose to pneumonia. Many lung cancers have a rich blood supply. The surface of the cancer may be fragile, leading to bleeding from the cancer into the airway. This blood may subsequently be coughed up. The main causes of any cancer include carcinogens (such as those in tobacco smoke), ionizing radiation, and viral infection. This exposure causes cumulative changes to the DNA in the tissue lining the bronchi of the lungs (the bronchial epithelium). As more tissue becomes damaged, eventually a cancer develops.

High-resolution X-ray computed tomography (CT) is the standard for pulmonary imaging. Depending on the scanner hardware, CT can provide high spatial and high temporal resolution, excellent contrast resolution for the pulmonary structures and surrounding anatomy, and the ability to gather a complete three-dimensional (3-D) volume of the human thorax in a single breath hold. Pulmonary CT images have been used for applications such as lung parenchyma density analysis, airway analysis, and lung and diaphragm mechanics analysis. A precursor to all of these quantitative analysis applications is lung segmentation. With the introduction of multislice spiral CT scanners, the number of volumetric studies of the lung is increasing and it is critical to develop fast, accurate algorithms that require minimal to no human interaction to identify the precise boundaries of the lung. A number of groups have developed techniques for computer assisted segmentation of pulmonary CT images manually traced boundaries were used to estimate regional gas and tissue volumes in the lungs of normal subjects. But manual methods are laborious. In an X-ray CT the natural contrast between the low-density lungs and the surrounding high-density chest wall can be used to guide image segmentation. More recently, Brown et al. provided a knowledge-based, automatic method to segment chest CT images. In their method, anatomic knowledge stored in a semantic network is used to guide low-level image processing routines. Segmentation algorithms for lung CTs in the literature are mostly pixelbased methods. In pixelbased methods, the first idea is to eliminate fat tissue and bones. As the lung parenchyma has a very lowdensity, it is composed of lowintensity pixels in the CT scan. This property is exploited to separate the two lungs from the surrounding tissue.
SCOPE: 1. Early detection of brain and lung cancer using MRI-image 2. Helps to identify cancer 3. Also used to find absence of tumor 4. Uses neural network feature

METHODOLOGY : 1. BRAIN CANCER: MRI Sample Images

Image

Preprocessing

Windowing

Feature

Extraction Suspicious Regions 1. Data Set

For the implementation of automated recognition system a data set collected from different source for various class of MRI image is considered. Figure shows the database considered for the implementation. The collected MRI images are categorized into four distinct classes with each as one type of cancer. The MRI scan are scanned and passed for implementation. 2. Image Segmentation
The first step is to segment the MRI image. Segmentation subdivides an image into its constituent parts of objects, the level to which this subdivision is carried depends on the problem being solved, that is, the segmentation should stop when the edge of the tumor is able to be detected.i.e. the main interest is to isolate the tumor from its background.The main problem in the edge detection process is that the cancer cells near the surface of the MRI is very fatty, thus appears very dark on the MRI, which is very confusing in the edge detection process. To overcome the problem, two steps were performed. First, histogram equalization has been applied to the image to enhance the gray level near the edge.

Second, thresholding the equalized image in order to obtain a binarized MRI with gray level 1 representing the cancer cells and gray level 0 representing the background.

3. Histogram Equalization The histogram of an image represents the relative frequency of occurrences of the various gray levels in the image. Histogram modeling techniques (e.g. histogram equalization) provide a sophisticated method for modifying the dynamic range and contrast of an image by altering that image such that its intensity histogram has a desired shape. Unlike contrast stretching, histogram modeling operators may employ non-linear and non-monotonic transfer functions to map between pixel intensity values in the input and output images. Histogram equalization employs a monotonic, non-linear mapping which re-assign the intensity values of pixels in the input image such that the output image contains a uniform distribution of intensities. 4.Thresholding In many vision applications, it is useful to be able to separate out the regions of the image corresponding to objects in which we are interested, from the regions of the image that correspond to background. Thresholding often provides an easy and convenient way to perform this segmentation on the basis of the different intensities or colors in the foreground and background regions of an image. The input to a Thresholding operation is typically a greyscale or color image. In the simplest implementation the output is a binary image representing the segmentation. Black pixels corresponds to background and white pixels correspond to foreground. In simple implementations, the segmentation is determined by a single parameter known as the intensity

threshold. In a single pass, each pixel in the image is compared with this threshold. If the pixels intensity is higher than the threshold, the pixel is set to white, in the output. If it is less than the threshold, it is set to black. Segmentation is accomplished by scanning the whole image pixel by pixel and labeling each pixel as object or background according to its binarized gray level. 5. Image Enhancement The fundamental enhancement needed in MRI is an increase in contrast. Contrast between the brain and the tumor region may be present on a MRI but below the threshold of human perception. Thus, to enhance contrast between the normal brain and tumor region, a sharpening filter is applied to the digitized MRI resulting in noticeable enhancement in image contrast. 6. Sharpening Filter Sharpening filters work by increasing contrast at edges to highlight fine detail or enhance detail that has been blurred. It seeks to emphasize changes. The most common sharpening filter uses a neighborhood of 3*3 pixel. For each output pixel it computes the weighted sum of the corresponding input pixel and its eight surrounding pixels. The weights are positive for the central pixel and negative for the surrounding pixels. By arranging the weights so that their sum is equal to one, the overall brightness of the image is unaffected. Weights can be adjusted as follows: -1 -1 -1 -1 0 -1 -1 -1 -1

7. Morphological operation
For the text region extraction, we use morphological operators and the logical operator to further remove the non-text regions. In text regions, vertical edges, Horizontal edges and diagonal edges are mingled together while they are distributed separately in non-text regions. Since text regions are composed of vertical edges, horizontal edges and diagonal edges, text regions can be determined to be the regions where those three kinds of edges are intermixed. Text edges are generally short and connected with each other in different orientation. Morphological dilation and Erosion operators are used to connect isolated candidate text edges in each detail component sub-band of the binary image. Figure 5 shows the Morphological operated scaled image.

8. Feature Extraction
The feature extraction extracts the features of importance for image recognition. The feature extracted gives the property of the text character, which can be used for training in the database. The obtained trained feature is compared with the test sample feature obtained and classified as one of the extracted character.

Texture features or more precisely, Gray Level Co-occurrence Matrix (GLCM) features are used to distinguish between normal and abnormal brain tumors. Five co-occurrence matrices are constructed in four spatial orientations horizontal, right diagonal, vertical and left diagonal (0, 45, 90 , and 135). A fifth matrix is constructed as the mean of the preceding four matrices.
Texture Features ( Gray Level Co-occurrence Matrix Features) From each co-occurrence matrix, a set of five-feautres are extracted in different orientations for the training of the neuro-fuzzy model.

Let P be the N*N co-occurrence matrix calculated for each sub-image, then the features as given by Byer are as follows :
1. Maximum Probability f1=max i,j p(i,j)

2. Contrast

3.Inverse Difference Moment (Homogeneity)

4. Angular Second Moment (ASM)

5.Dissimilarity

6.Grey Level Co-occurrence Mean (GLCM)

7. Variance:

8.Correlation Coefficient

Where

9. Entropy

9. Feature Selection Feature selection concerns the reduction of the dimensionality of the pattern space and the identification of features that contain most of the essential information needed for discriminating between normal and abnormal cases. Selection of efficient features can reduce significantly the difficulty of the classifier design. Therefore feature selection based on the correlation coefficient between features is performed. The correlation matrix was calculated for the set of 9 texture features for both normal and abnormal spaces.
Any two features with correlation coefficient that exceeds 0.9 in both spaces can be combined together and thought as one feature reducing the dimensionality of the feature space by one. Therefore the maximum probability and contrast can be removed and the numbers of features are reduced to seven features.

10. Neuro-Fuzzy Classifier

WORKING 1. Brain image is read and displayed 2. Selected image is loaded and expectation maximization segmentation algorithm is applied. 3.Nine distinct invariant features with calculation of minimum distance for the prediction of cancer are used to predict tumor in given MRI image 4. Histogram employs a motonic ,non-linear mapping which re-assign the intensity values of pixels in the input image such that output image contains a uniform distribution of intensities 5. Thresholding is used to separate out the regions of the image corresponding to objects in which we are interested from regions of image that corresponds to the background. 6. Neural classification is used for recognition.

LUNG CANCER DETECTION:

We assume that the input data is 2-D thoracic CT, which is in JPEG format. The data set should include thorax from the trachea to below the diaphragm. In the CT image data, air will appear with a mean intensity of approximately -1000 Hounsfield units (HU), most lung tissue will be in the range of -910 HU to -500 HU, while the chest wall, blood and bone will be much more dense (above -500 HU). As shown in Fig., this method consists of five main steps: an extraction step to identify the lungs, a separation step to separate the right and left lungs, smoothing step to smooth the lung boundaries, feature extraction step to extract texture features and last step is to classify the lung diseases using Back Propagation neural network and Feed forward cascade neural network. A. Lung Extraction 1) Fuzzy C-means Clustering Fuzzy c-means (FCM) is a method of clustering which allows one piece of data to belong to two or more clusters. Fuzzy partitioning is carried out through an iterative optimization of the objective function. This procedure converges to a local minimum. Fuzzy c-Means clustering is used to segment the lungs. The maximum Number of iterations specified is 100. The

termination criterion is fixed as .00001.The pixels of the input image are divided into 2 clusters. The first cluster includes pixels in the lungs (body pixel). The second cluster includes pixels in the thoracic CT other than lungs (non-body pixel). The pixels corresponding to the first cluster is determined. 2) Background Removal Rarely holes can appear near the border of the parenchyma. To fill these holes and glue different parts of a same lung together, a rolling ball operator is applied. Rolling ball operator is in fact a morphological closing of the region followed by hole filling. A hole is a set of background pixels that cannot be reached by filling in the background from the edge of the image. As the air around the body has a very similar intensity to the lungs it will not be discarded by the thresholding, so it has to be removed. Further steps are performed to improve the result. Back ground pixels are identified as follows: They are non-body pixels and pixels connected to the border. Thus every connected region of non-body pixel that touches the border is considered as background and discarded. B. Lung Separation The segmented lungs must be identified and separated. The region is simply cut vertically in the middle. And the resultant image is multiplied with the original image to get the right and left lungs. C. Post-Processing The morphological operations such as opening followed by closing are again applied after separating the lungs to smooth it. The morphological operations is defined by f * b = (f b ) b where f is the target image, b is the structuring element, means morphological closing, means morphological opening the images obtained after applying disc based structuring element are multiplied with the original image to get the segmented lungs. D. Feature Extraction 1) Covariance Features

Auto-correlation coefficients reflect the inter-pixel correlation within an image. It is a Mean removed version also called Modified Auto Covariance features. Let the given image has size M*N. The modified auto-covariance coefficients is given by Where (m, n) = 1 - (m, n) / (0,0) (m, n) = 1 / (M - m)(N - n)* M-1-m N-1-n (f(x,y) mn)(f(x+m,y+n)-mn) x=0 y=0 m = 0 .. a n = 0 .. b a = 1 .. 4 b = 1 .. 4 E. Classification Of Lung Diseases 1) Cascade Forward back propagation The commonest type of artificial neural network consists of three groups, or layers, of units: a layer of "input" units is connected to a layer of "hidden" units, which is connected to a layer of "output" units. The activity of the input units represents the raw information that is fed into the network. The activity of each hidden unit is determined by the activities of the input units and the weights on the connections between the input and the hidden units. The behavior of the output units depends on the activity of the hidden units and the weights between the hidden and output units. The behavior of an ANN (Artificial Neural Network) depends on both the weights and the input-output function (transfer function) that is specified for the units. Cascade-forward is a feed forward network with the exception that it starts with a minimal number of hidden layer neurons. During back propagation additional neurons are added to hidden layer as needed to improve classification. CFBP is just like BPN. It is also using back propagation algorithm for updating weights. But main symptom of this network is that each layer neurons closely related to all previous layer neurons. Training process towards these networks is iterative. When the error between desired value and predicted value became minimum, training process towards stability.

CONCLUSION

It is observed that the system result in better classification during the recognition process. The considerable iteration time and the accuracy level is found to be about 50-60% improved in recognition compared to the existing neuro classifier.