Introducing hybrid k-means, RLS learning algorithm in RBF network for low computational brain MRI multi-classification

In this paper we study the performance of a Radial Basis Function (RBF) network with hybrid learning schemes on classification of multiple brain disease based on MR (Magnetic resonance) Image processing. The proposed method is shown to be superior to all previously presented classifiers including supporting vector machines (SVM), K-nearest neighbourhood (KNN) and simple RBF networks. Also the most efficient feature extraction and reduction methods were chosen based on computational complexity. Our aim in this paper along with the complexity reduction is the introduction of hybrid ‘k-means, RLS’ classifier to the sophisticated field of MRI classification. Also an improvement is made in this paper to the multiple classification of brain disease in 10 class scenario.

Keywords:

Brain MRI , image classification , K-means algorithm , adaptive algorithms , Radial basis function