Brain Tumor Diagnosis Using a Hybrid Method of BoVW, LBP, and Neural Networks

Brain tumor is one of the most deadly diseases today. Early identification is the key in treatment and automated method of image classification can help in this process. But, there are significant obstacles in precisely identifying brain tumors and increasing the classification accuracy for different types of tumors. In order to address the problem, the work suggests a hybrid model for effectively detecting and classifying brain tumors. First, the obtained pictures are preprocessed with resizing, application of sharpening filter and final stage of contrast enhancement. Further preprocessing in carried out using the K-mean clustering algorithm Next, Feature extraction is done by using the BoVW together with LBP which deal with the overall and region-specific texture features respectively. Following that, the seven statistical features are extracted, and then they are utilized as input to the artificial neural network classifier in order to obtain the classification of the glioma, meningioma, and pituitary tumor. The training function ‘Levenberg-Marquardt’ is also used in this work. This method provides 99.0 % of accuracy on Dataset I and 99.2% accuracy on Dataset II. Moreover, in relation to other modern methods, the selected seven statistical features employed in this study are relatively limited in demonstrating this performance.