Piecewise Fuzzy C-Means Clustering and Deep Convolutional Neural Network for automatic brain tumour classification using MRI images
Magnetic Resonance Imaging (MRI) provides lots of information about human soft tissue, which is employed in radiology for the diagnosis of brain tumour. Accuracy and early detection are the major concerns in tumour diagnosis. This paper develops an automatic brain tumour classification method using MRI images based on the Piecewise Fuzzy C-Means Clustering (pifCM) and Deep Convolutional Neural Network (Deep CNN). Initially, the contriteness of the input MRI image is enhanced through pre-processing the image using the piecewise fuzzy c-means clustering method. The next step is feature extraction in which the texture features and statistical features are extracted using Local directional pattern (LDP), wavelet transform, principal component analysis (PCA), entropy, and mean. Finally, the tumors are classified using Exponential cuckoo-based deep convolutional Neural Network (Exponential cuckoo-based DCNN) classifier. The simulation of the proposed method of tumor classification is done using BRATS and SIMBRATS database and the performance obtained by the proposed is compared with several state-of-art techniques. The simulated results show higher accuracy of 0.8711 and minimal Mean Square Error (MSE) of 0.0197when compared with the existing methods.