Performance Comparison of Adaptive Neural Networks and Adaptive Neuro-Fuzzy Inference System in Brain Cancer Classification

Brain tumors are amongst the top death-leading health conditions worldwide. Biopsy is the most accurate procedure that determines the brain tumor type whether it is malignant or benign. However, biopsy may not be applicable for some patients with brain cancer (BCa) and could be life-threatening. In this paper, an intelligent diagnosticimage-based systems are implemented to assist physicians in making diagnostic decisions about the BCa type without biopsy procedures. A combined method of artificial intelligent systems and MRI image segmentation is proposed as a tumor classification tool. This study employs image filtration and segmentation on a region of interest (ROI) of an MRI image. Then, extract accurate statistical features are fed into four artificial intelligent (AI) systems: Adaptive neuro-fuzzy inference system(ANFIS), Elamn Neural Network (Elman NN), Nonlinear Auto Regressive with exogenous neural networks (NARXNN), and feed forward NN. The four AI classifiers are investigated and tested on 107 patients with brain tumors. The data base of the brain tumor images used in this study contains both malignant and benign cancers. The performance of the four intelligent tumor classifiers is evaluated. It is found that the NARX NN shows best performance with a classification accuracy of 99.1%. The achieved accuracy level is superior and could be very helpful in clinical purposes. (C) 2014 Jordan Journal of Mechanical and Industrial Engineering. All rights reserved