Tumor Detection in Brain MRI Image Using Template based K-means and Fuzzy C-means Clustering Algorithm

ABSTRACT:

This paper presents a robust segmentation method which is the integration of Template based K-means and modified Fuzzy C-means (TKFCM) clustering algorithm that, reduces operators and equipment error. In this method, the template is selected based on convolution between gray level intensity in small portion of brain image, and brain tumor image. K-means algorithm is to emphasized initial segmentation through the proper selection of template. Updated membership is obtained through distances from cluster centroid to cluster data points, until it reaches to its best. This Euclidian distance depends upon the different features i.e. intensity, entropy, contrast, dissimilarity and homogeneity of coarse image, which was depended only on similarity in conventional FCM. Then, on the basis of updated membership and automatic cluster selection, a sharp segmented image is obtained with red marked tumor from modified FCM technique. The small deviation of gray level intensity of normal and abnormal tissue is detected through TKFCM. The performances of TKFCM method is analyzed through neural network provide a better regression and least error. The performance parameters show relevant results which are effective in detecting tumor in multiple intensity based brain MRI image.

PROJECT OUTPUT VIDEO:

EXISTING SYSTEM:

• In the clustering technique, Fuzzy c-means (FCM) clustering and expectation–maximization (EM) algorithms are being the most widely used methods for clustering. The applications of the EM algorithm to brain MR image segmentation and a common disadvantage of EM algorithms are reported Wells et al..

• There is a FCM algorithm which contacts with a knowledge based classification and tissue labeling used Li et al.. Firstly, this FCM method segments MR brain images, and then introduces an expert system to locate a landmark tissue by matching them with a prior model. An ANN is used Hall et al., and compared the performance with FCM for segmenting brain MR images.

• Conventional FCM Pham et al. has limitation of noise sensitivity and imperfection to the abnormality of brain e.g. tumor, edema, and cyst. Although kmeans segmentation is noise immune, but it is prerequisite of this method that there should be perfect thresholding, which is quite hard for complex brain structure.

DISADVANTAGES OF EXISTING SYSTEM:

• An appropriate segmentation of brain MRI image is apparent for detecting abnormality in brain. As brain comprehends complicated structure so segmentation of MRI image obliges good care and should be precise.

• Due to structure complexity of brain tissue, proper threshold value is very hard to achieve. The main drawback of thresholding is that, it cannot be applicable for multiple channel images. In addition, it does not provide spatial characteristics, which causes it to be sensitive to noise as well as inhomogeneity intensity.

PROPOSED SYSTEM:

• In this paper, we have proposed a different technique for tumor detection from brain MRI image, based on the combination of template based K-means and modified FCM (TKFCM). It will reduce the problem of template or gray level selection and noise sensitivity accustomed to FCM and, Region growing.

• In this paper, template based k-means extension has applied through pixel intensity positioning. Very little bit of pixel intensity is not avoided as the template is selected based on the number of gray level to be used and the coarse image. FCM algorithm is modified on the basis of updated membership and number of clusters for the filtered image.

ADVANTAGES OF PROPOSED SYSTEM:

• The membership values are updated based on the features such as intensity, entropy, contrast, and homogeneity of the MRI brain image. By choosing right membership input & output variables along with adjusted number of feature and cluster the detection of tumor is completed nicely.

• The whole performance is analyzed through neural network, which is highly accepted in the segmentation field. It is done by preparing train and target data for the relevant image through the network with feed forward back-propagation algorithm.

MODULES:

1. Image Acquisition

2. Image Enhancement

3. Segmentation

MODULE DESCRIPTION:

1. Image Acquisition:

• Images are acquired from Gallery.

2. Preprocessing:

• The objective of the preprocessing phase is to apply possible image enhancement techniques to obtain the required visual quality of the brain tumor image.

• We enhance the brain tumor image by using imadjust function.

3. Segmentation:

• After preprocessing stage, we propose segmentation process.

• In that, brain tumor region was segmented using template K-means clustering and modified C-means clustering.

• The proposed algorithm is integration of the k-means and fuzzy c-means with some modification.

• The template is added along with the conventional k-means, which is identified by the temper or gray level intensity in the brain image.

• Besides the fuzzy c-means membership and Euclidian distance is modified by the image features (Energy, Contrast, Correlation, and Homogeneity).

• Template based k-means and modified fuzzy c-means clustering algorithm for segmentation can be written in equation as below:

SYSTEM REQUIREMENTS:

HARDWARE REQUIREMENTS:

• System           : Pentium Dual Core.

• Hard Disk      : 120 GB.

• Monitor          : 15’’ LED

• Input Devices : Keyboard, Mouse

• Ram                : 1GB.

SOFTWARE REQUIREMENTS:

• Operating system : Windows 7.

• Coding Language : MATLAB

• Tool                       : MATLAB R2013A

REFERENCE:

Rasel Ahmmed, Md. Foisal Hossain, “Tumor Detection in Brain MRI Image Using

Template based K-means and Fuzzy C-means Clustering Algorithm”, IEEE CONFERENCE, 2016.