MRI IMAGE ENHANCEMENT DETECTING AND RECOGNITION ABNORMAL IN THE BRAIN USING NEURAL NETWORKS COMBINED ANISOTROPIC DIFFUSION FILTER

Abstract

Anomaly recognition is one of the most important and difficult tasks in the field of medical image processing because manual classification can lead to inaccurate predictions and diagnoses. Moreover, nowadays in hospitals there is a large amount of medical imaging data that needs to be processed, which makes this task even more important and urgent. Abnormalities in the brain such as brain tumors are highly variable in appearance and similarity between the tumor and normal tissues and thus it becomes more difficult to extract these tumor regions from imaging. In this paper, we have proposed a method to improve the quality of 2D Magnetic Resonance Brain Image (MRI) by neural network combined with anisotropic filtering as an important pre-processing solution for detecting and recognition abnormalities in the brain. The main purpose of the paper is to optimize the MRI image processing solution in denoising, detecting and recognition anomalies with higher accuracy. In this method, anisotropic filtering is a technique that aims to reduce image noise without losing important parts of the image content, usually edges, lines or other details important to the image representation. The neural network is then used to remove residual noise and enhance the image to distinguish between normal and abnormal pixels, based on texture features and on statistics, which gives better performance and higher accuracy than traditional methods. By analyzing and calculating results of experimentally processed image quality parameters, we will prove that the proposed method is superior to some traditional methods such as Gaussian filter, filter wiener.