Exponential cuckoo search algorithm to Radial Basis Neural Network for automatic classification in MRI images

Abstract

Generally, Magnetic Resonance Imaging (MRI) is utilised in radiology for diagnose the anatomy and the physiological processes of the body. Nowadays, the classification of tumour region plays a vital role in MRI brain imaging technique. Due to variance and complexity of tumours, the classification and segmentation of tumour are burdensome in MRI brain images. This paper proposes a Radial Basis Neural Network (RBNN) based on exponential cuckoo search algorithm for the automatic classification of tumour in the brain. Initially, the fuzzy c-means clustering is employed to the segmentation for the detection of tumour region. Then, the features are extracted from the tumour and nontumour regions that are concatenated to generate the feature vector. These features are applied to the proposed classifier RBNN. This classifier requires the optimal cluster centre which is iteratively evaluated by the newly proposed exponential cuckoo search algorithm. Thus, the classifier classifies the tumour and nontumour images and also determines the severity of tumour. The proposed system is analysed for the evaluation metrics, such as segmentation accuracy, MSE and accuracy. Thus, the proposed system attains the higher accuracy 89% which ensures, the better classification of MRI brain image.

Keywords:

• MRI images
• segmentation
• feature extraction
• RBNN
• cuckoo search algorithm
• classification