An Efficient Method for Automatic Classification of Brain MRI using Feature Selectionand Modified Probabilistic Neural Network
Bobbala Sreedevi1, T. Anil Kumar2, K. Kishan Rao3
1Bobbala Sreedevi*, Research Scholar in JNTUH and Working as Assistant Professor, Dept. of ECE, Vaagdevi College of Engineering, Warangal Telangana, India.
2Dr.T.Anil Kumar, Professor, Department of ECE, CMR Institute of Technology, Hyderabad, Telangana, India.
3Dr. K. Kishan Rao, Professor, Department of ECE, Sreenidhi Institute of Science and Technology, Hyderabad, Telangana, India.
Manuscript received on November 13, 2019. | Revised Manuscript received on 24 November, 2019. | Manuscript published on December 10, 2019. | PP: 976-981 | Volume-9 Issue-2, December 2019. | Retrieval Number: I8220078919/2019©BEIESP | DOI: 10.35940/ijitee.I8220.129219
Open Access | Ethics and Policies | Cite | Mendeley | Indexing and Abstracting
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC-BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: In the brain tumor MRI images, the identification, segmentation and detection of the infectious area is a tedious and lengthy task. As segmentation is called intensity inhomogeneity by an intrinsic object. In this paper we suggest an energy efficient minimization technique for joint domain assessment and segmentation of MR images called multiplicative intrinsic component optimization (MICO). In this work, we focused on quicker implementation with a robust removal of gray-level co-occurrence matrix (GLCM). Optimal texture characteristics are obtained by the Spatial Gray Dependence (SGLDM) technique from ordinary and tumor areas. With very large feature sets, the choice of features is redundant because the precision frequently worsens without choice of features. However, when only the feature selection is used, the precision of classification is significantly improved. However, by reducing the time needed for classification computations and improving classification precision by removing redundant, false or incorrect characteristics. A fresh function choice and weighting technique, supported by the decomposition developmental multi-objective algorithm, are provided in this work. These characteristics are provided for the MPNN classification. Modified probabilistic neural network (MPNN) classification was used in brain MRI images for training and testing for precision in tumor identification. The simulation findings accomplished almost 98% precision in the identification of ordinary and abnormal tissue from brain MR images showing the efficiency of the method suggested.
Keywords: Brain Segmentation, Intensity Inhomogeneity, Texture Features, Probabilistic Neural Networks, Bias field Estimation
Scope of the Article: Classification