Multilayer Perceptron for Brain Image Classification using Stationary Wavelets

The majority of individuals who are diagnosed with malignant brain tumors do not survive, despite recent developments in the methods of neurosurgery and treatment. The diagnosis of high-grade gliomas at an earlier stage is still challenging. In recent years, Magnetic Resonance Imaging (MRI) has emerged as one of the most useful research methods for studying the human brain. An efficient MRI brain image classification (MBIC) system is developed in this work to predict abnormality in brain at an earlier stage. At first, the MRI images are represented in stationary wavelet domain. The truncated Singular Value Decomposition (t-SVD), a linear dimensionality algorithm is used for reducing the feature vector obtained from stationary wavelets. Then, Multi-Layer Perceptron (MLP) is employed for the classification of MRI images from REpository of Molecular BRAin Neoplasia DaTa (REMBRANDT) database. Results proved that the proposed MBIC system achieves 98.8% overall accuracy with 98.6% sensitivity and 99% specificity.