الفهرس 
AbstractOnly 14 pages are availabe for public view
 |  | from | 199 |  |  |
| | | from | 199 | | |
Abstract
Brain tumor is one of the most harmful disease ,and has affected majority of people. The proposed non invasive method involves type identification of tumor. DICOM MRI images is efficient but its low contrast ,so the image firstly converted to grayscale image to improve visual appearance ,then we need to find brain tumor by histogram analysis after that contrast stretching for enhancement of gray level of these images have been done in order to increase the dynamic range. The next stage involves MDADF denoisiong method to remove noise of MRI. Using (MDADF) as it preserves fine structures and reduces correlated noises with PSNR = and SSIM= when adding Gaussian nose with α=30.MDADF has high PSNR=27.231 and SSIM=0.13 when adding high noise =18,put when using ADF with PSNR=25.7 The third stage is skull stripping .This step is important part in CAD.The algorithm was validated also by comparing its performance with exiting automated skull stripping method. It outperform these method due to its simplicity and speed with Sensitivity=0.99 and Specificity =0.99 The fourth stage include an perfected segmentation method founded on Neutrosophic sets (NS) and Modified Non local Fuzzy c-mean clustering(MNLFCM) is proposed .The planned method denoted as NS- MNLFCM-MLS and compared with additional paper using Jaccard and Dice Constant. The investigational outcomes prove that the offered method is fewer subtle to din and does better on MRI brain appearance .The Dice=0.9937 ,Sensitivity=0.9412 and Specificity=0.9926 this results are more then another methods IFCM and IVIFCM The fifth stage using hybrid FE is suggested as HOG(Histogram of orainted),LBP(local binary pattern), SIFT(Scale-invariant feature transform) , gray level co_occurance matrix (GLCM),counting label occurance matrix (CLOM),Modified counting label occurance matrix(C2LOM) and Anisotropic diffusion wavelet transform (ADWT). Feature selection (FS) technique as gray wolf optimization (GWO) technique this technique provides good outcome for great occurrence using to reduce large num of feature used.This investigation result demonstrate that the suggested hybrid procedure of FE technique provides better correctness than early procedures once experienced with classifiers as Support vector machine (SVM),Linear vector quantization (LVQ),Self organization map (SOM) and our proposed method called Multi layer perceptron with Dragonfly optimization method (MLPDO). The desired outcomes were stated as malignant and benign. Next decide that the tumor is malignant, we can decide likewise the level of malignant and benign tumor if it is low grade or high grade.By using MLPDO we get high accuracy rate robust result with low mean square error = 0.0012 ±7.4498e-05.