Automated Segmentation and Detection of Brain Tumor from MRI

To increases the survival rate of the brain tumor patients and to have a improved treatment technique in medical image processing, brain tumor segmentation is essential method of diagnosis. The early and correct diagnosis of brain tumors plays an important role. Magnetic Resonance Imaging (MRI) technique is the most popular non-invasive technique; in these days imaging of biological structures by MRI is a common investigating procedure. For cancer diagnosis the brain tumors segmentation can be done manually from MRI, which gives the poor level of accuracy and identification. The classification of abnormalities is not predictable and straightforward but it is a time consuming task for physician. Nowadays, the issue of automatic segmentation and analysis of brain tumours are major research area. However the detection of tumor is a challenging task since tumor possesses complex characteristics in appearance and boundaries. In order to produce a completely automated segmentation method like the KG (knowledge-guided) technique which encrypts the information of the pixel intensity and spatial relationships in the images. The k NN classifier under the learned optimal distance metrics is used to determine the possibility of each pixel belonging to the foreground (tumour) and the back ground. The paper presents semi-automatic segmentation method by using CNN (Convolutional neural networks) on the basis of individual statistical information and population, to segment brain tumours early, increase the correct rate and minimize error rate. The experimental result of proposed method demonstrates the robustness for brain tumor segmentation. It shows improved result for classification of Brain tumor from MRI of brain than k-NN Classifier.