Design of Deep Neural Architecture for Brain Cancer Classification Using Pyramid Design

Abstract

Listen

Translate article

In this study, Deep Neural Network (DNN) with Pyramid Design (PD) is developed for brain image classification. The main objective is to design a non-invasive procedure for diagnosing brain cancer using deep learning. The signal intensities obtained from the brain tissues using Magnetic Resonance Imaging (MRI) can be used for effective treatment. A pyramid consists of a predefined number of convolution layers and a max pool layer to abstract features. In the DNN-PD approach, the pyramid is stacked with an increasing number of convolution layers and convolution filters from one pyramid to the next pyramid and so on. The stacking is employed recursively to get more accurate results. It shows that the proposed architecture gives efficient results with 98.5% accuracy using REpository of Molecular BRAin Neoplasia DaTa (REMBRANDT) database. The diagnostic results of the proposed Deep Neural architecture with the expert’s analysis can reduce the biopsies, a dangerous procedure with a great deal of pain.