Presenting a method based on image processing using data mining techniques to classify brain tumors in MRI images

Abstract

Accurate and immediate diagnosis of brain tumors is essential to implement effective treatment of this disease. The choice of treatment depends on the level of the tumor at the time of diagnosis, the type of pathology and the degree of the tumor. Computer-aided diagnostic techniques have helped neurologists in a variety of ways. Recent advances in computer-aided medical diagnosis have led to improved performance with the advent of the concept of deep learning. In this study, in order to analyze and diagnose brain tumors using image processing; The use of convolution neural network algorithm method for processing MRI images of the brain has been suggested. The result shows that our approach to different types of noise and their mixtures is stronger and reduces the probability of failure in several cases. Therefore, the convergence criterion always used by the FCM algorithm and its variants does not seem to be sufficient to guarantee the improvement of the final results compared to the initial or intermediate results. The results of the proposed approach show that even repetition for the first time is repetitive enough to make the right decision with high probability. The main drawbacks of different versions of the modified FCM are the use of blind filtering, such as mean (weight) with spatial constraints, which may lead to the introduction of a new class of actual classes in the area leading to inappropriate classification. We have shown that our consistency improves the rules of results with little computational effort.