A hybrid approach based on multiple Eigenvalues selection (MES) for the automated grading of a brain tumor using MRI

•The segmentation, identification, and classification of brain tumor using magnetic resonance images are essential for making a correct diagnosis. Computer aided technology has therefore been developed to computerize these procedures.•Finding a fully automated system for detection, segmentation and grading of the brain tumor is so useful in early detection of a brain tumor.•A large number of inputs to any system almost can cause some challenges such as over-fitting, or high computational complexity. Therefor it is important to find some way to decrease inputs and in the same time improve the outputs.•Although, many researches have been implemented regarding brain tumor segmentation and classification using MRI, this field of study is still an open area for investigation. Background and objective:The manual segmentation, identification, and classification of brain tumor using magnetic resonance (MR) images are essential for making a correct diagnosis. It is, however, an exhausting and time consuming task performed by clinical experts and the accuracy of the results is subject to their point of view. Computer aided technology has therefore been developed to computerize these procedures. Methods:In order to improve the outcomes and decrease the complications involved in the process of analysing medical images, this study has investigated several methods. These include: a Local Difference in Intensity - Means (LDI-Means) based brain tumor segmentation, Mutual Information (MI) based feature selection, Singular Value Decomposition (SVD) based dimensionality reduction, and both Support Vector Machine (SVM) and Multi-Layer Perceptron (MLP) based brain tumor classification. Also, this study has presented a new method named Multiple Eigenvalues Selection (MES) to choose the most meaningful features as inputs to the classifiers. This combination between unsupervised and supervised techniques formed an effective system for the grading of brain glioma. Results:The experimental results of the proposed method showed an excellent performance in terms of accuracy, recall, specificity, precision, and error rate. They are 91.02%,86.52%, 94.26%, 87.07%, and 0.0897 respectively. Conclusion:The obtained results prove the significance and effectiveness of the proposed method in comparison to other state-of-the-art techniques and it can have in the contribution to an early diagnosis of brain glioma.