ANN-Based Brain Tumor Classification: Performance Analysis Using K-Means and FCM Clustering With Various Training Functions

Abstract

Brain tumor is a serious and complicated disease for human beings and it is obligatory to detect it at an early stage. Classifying the tumor types from MRI images without skilled radiologists is challenging. Automatic brain tumor classification is an efficient way as it can provide accurate results without wasting time. Classifying brain tumors flawlessly is the paramount theme of this chapter. Two algorithms are proposed in this chapter for MRI brain tumor classification by comparing two major clustering algorithms with artificial neural network (ANN). For algorithm 1, MRI images are pre-processed by reseized and sharpening filter firstly. Secondly, images are segmented by k-means clustering. Thirdly, 2D discrete wavelet transform is utilized to extract features and those features are compressed through principal component analysis. Finally, compressed features are applied for ANN training, validation and testing. This feed-forward network is trained by nine training functions separately to observe the accuracy variation. For algorithm 2, fuzzy c-means clustering is used instead of k-means and other procedures remain exactly the same as for the previous method. The proposed first approach provides 98.0% classification accuracy and second approach provides 99.8% classification accuracy with Levenberg-Marquardt function. Performance evaluation of the proposed work concludes that the second approach shows better classification accuracy than the first one and Levenberg-Marquardt training function provides the best accuracy than other training functions. The best result of proposed work is best among various existing works.