Brain Tumor Segmentation and Identification Using Particle Imperialist Deep Convolutional Neural Network in MRI Images

• Published in: International journal of interactive multimedia and artificial intelligence Vol. 7; no. 7; pp. 38 - 47
• Main Authors: Khemchandani, Maahi Amit, Jadhav, Shivajirao Manikrao, Iyer, B.R
• Format: Journal Article
• Language: English
• Published: IMAI Software 01.12.2022; Universidad Internacional de La Rioja (UNIR)
• Subjects: Algorithms; Brain tumors; Cancer; deep belief network; Imperialism; local neighborhood structure; Magnetic resonance imaging; Medical imaging equipment; Neural networks; scattering transform; Technology application; tumor; tumor characterization; Tumor Deep Belief Network Local Neighborhood Structure Scattering Transform Tumor Characterization
• ISSN: 1989-1660; 1989-1660
• DOI: 10.9781/ijimai.2022.10.006

For the past few years, segmentation for medical applications using Magnetic Resonance (MR) images is concentrated. Segmentation of Brain tumors using MRI paves an effective platform to plan the treatment and diagnosis of tumors. Thus, segmentation is necessary to be improved, for a novel framework. The Particle Imperialist Deep Convolutional Neural Network (Pi-Deep CNN) suggested framework is intended to address the problems with segmenting and categorizing the brain tumor. Using the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) Algorithm, the input MRI brain image is segmented, and then features are extracted using the Scatter Local Neighborhood Structure (SLNS) descriptor. Combining the scattering transform and the Local Neighborhood Structure (LNS) descriptor yields the proposed descriptor. A suggested Particle Imperialist algorithm-trained Deep CNN is then used to achieve the tumor-level classification. Different levels of the tumor are classified by the classifier, including Normal without tumor, Abnormal, Malignant tumor, and Non-malignant tumor. The cell is identified as a tumor cell and is subjected to additional diagnostics, with the exception of the normal cells that are tumor-free. The proposed method obtained a maximum accuracy of 0.965 during the experimentation utilizing the BRATS database and performance measures. KEYWORDS Brain Tumor Segmentation, Deep Belief Network, LNS Descriptor, Scattering Transform, Tumor Level.