Original Articles
Leena Nangai V
Abstract
Nowadays, Brain tumor is one of the most dangerous diseases occurring commonly among human beings. By using Computed Tomography Imaging and Magnetic Resonance Imaging experts can diagnose a brain tumor. Computed Tomography (CT) imaging brain tumor images Classification is a complicated process due to the complexity and variance of brain tumors. A numerous research work has to be done on the Classification system for accurate image classification. However, the most of the existing works still could not achieve the automatic image classification. This research work introduces a technique from machine learning approach to effective and automatic image classification. The Hyperbolic Hopfield Neural Network (HHNN) technique is used to classify the given brain tumor query image. The proposed framework consists of three stages, namely, Segmentation, Texture extraction, and classification. In first stage, the bilateral filter is applied to remove the noises in obtained CT image. Then the sample filtered images are segmented by using Enhanced Markov Random Fields Approach (EMRF). In the second stage, the features are extracted from the segmented images using texture descriptor. Final stage is classification where the extracted features are trained using the proposed HHNN with hyperbolic Hopfield tangent activation function. At last, a query image is classified effectively using the trained features of HHNN. Finally, we conduct set of experiments to prove our proposed approach is better than other approaches.