MRI (Magnetic resonance Imaging) brain tumor images Classification is a difficult task due to the variance and complexity of tumors. This paper proposed techniques to classify the MR human brain images. The proposed classification technique consists of three stages, namely, pre-processing, feature extraction and selection, and classification. Features are extracted by using the gray level co-occurrence matrices and the gray level run length matrices (GLCM & GLRLM), 18 features were determined from the image, then selected the most important features that saved to the database. In the final stage, the classifier based on probabilistic neural network (PNN) have been used to classify MRI brain images, the proposed algorithm is trained with 50 images of (Sarcoma, Anaplastic Astrocytoma, Meningioma, and Benign) and tested with 65 images. The accuracy of this method was up to 98%.
Figure 6: PNN Architecture& Training Parameters
PNN classification evaluationThe evalution of the perofmance and accuracy of the proposed PNN is dpending on the following: 1. The first method is to test the network on the same sample features that trained on, then show the network how to classify them. Figure 7 shows that the network classifies each sample feature exactly to there class type.