In the field of medical imaging, brain segmentation and detection using MRI has become one of the most important areas of research and motivation. In any manual division of brain tumors and their use using magnetic resonance imaging of the brain, it forms a large part of the human intervention to break the patient and detect it, both difficult and has great internal and external evaluation and stress. Thus, there is a great demand for automatic detection and dissection of brain tumors using MRI images to overcome manual fragmentation. So in previous and present years, a number of methods have been proposed by researchers. But the complete automated system has not yet been developed due to problems of accuracy and time. So, this thesis provides a review of the methods and techniques that used to detect and segment brain tumor through MRI segmentation. Finally, the thesis concludes with one (hybrid methods, improving algorithm ) that show high accuracy in detecting brain tumor. This paper propose a new clustering algorithm, which relies on the differences between the contrast levels of the Tumor in the MRI. In the following, the thesis was developed in the k-mean algorithm and made less time consuming and more accurate in determining the tumor than the traditional algorithm and make her non-random in the selection of cluster number through a number similar to the value of the adjustment, which was chosen to match all images in the data set and choose the values of the initially clusters according to the laws developed by the thesis and this makes her possess the stability of the result of diagnosis of the presence or absence of the tumor, they appear in the same mass even when again attempt to re-divide the clusters, the suspicious body will appear in the same cluster that appeared in the previous time and does not take another cluster to find it. This is contrary to the traditional k-mean algorithm also improving algorithm depends on the activation function instead law distance that was using in traditional algorithm k-mean. The assignment of thesis consisted of some preprocessing steps like noise removal or reduce and enhanced the MRI by using enhancement techniques such as adaptive wiener filtering, haar wavelet transforms and morphological image operation. The main contribution of this thesis is additional to improve the k-means is combining the improved k-means clustering approach with fuzzy c-mean based on a new image fusion approach. This method depends first on the wavelet transform to images and then rely on the sub-bands resulting from the wavelet transform instead of the total range of the image adopted in the traditional way. Then use three different fusion techniques together at a different level in the fusion and use different levels of wavelet transform instead of a single level to achieve high image fusion and tumor identification. The experimental results of this thesis show that thesis approach have get better accuracy , less time consuming and less rate of errors. The system relies on the GT of the images of MRI after the mark has been placed on the entire data set by a specialized surgical operation with high efficiency.
Volume 11 | 01-Special Issue
Pages: 446-469