Detection and Segmentation of Meningioma Tumors Using the Proposed MENCNN Model

Abstract

This paper develops a Meningioma Detection and Segmentation System (MDSS) using the proposed Meningioma Convolutional Neural Network (MENCNN) classifier. The main objective of this paper is to detect and locate the meningioma brain tumors using the proposed deep learning structure and segmentation algorithm. This proposed MDSS is designed with preprocessing of meningioma and healthy brain MRI images, feature computations and feature classification through the proposed MENCNN classifier and Meningioma Segmentation Algorithm. The noises in both meningioma and healthy brain images are removed using Mean Adaptive Filter (MAF) and the meningioma features are computed from the noise removed image. These meningioma features are classified by the proposed MENCNN classifier in order to obtain the classification results as either meningioma or healthy brain image. Finally, Meningioma Segmentation Algorithm (MSA) is proposed in this research work to segment the pixels belonging to the meningioma region. The proposed MDSS approach obtains 96.46% MSI, 97.75% MSR and 97.6% MSA on the set of meningioma images in Nanfang dataset. The proposed MDSS approach obtains 97.76% MSI, 98.03% MSR and 97.81% MSA on the set of meningioma images in Kaggle dataset.