Research article

A RAPID COSINE SWARM OPTIMIZATION (RSCSO) – EXTREME LEARNING MACHINE (ELM) FOR AN AUTOMATED BRAIN TUMOR DIAGNOSIS SYSTEM

Sayeedakhanum Pathan1, Dr.Savadam Balaji2

Online First: December 22, 2022


For detecting tumor and non-tumor cells in the brain and evaluating cell levels, an efficient classification and segmentation of brain tumors is the intriguing aspect of the categories. Based on their experiences, the classification and segmentation of brain tumors were developed by many researchers in the medical field. Yet, traditional works have the critical problems of over-segmentation, overfitting, high time consumption, and error rate. Therefore, the proposed work motivates to development of an automated and highly efficient diagnosis framework for brain tumor detection and segmentation. Here, an iterative preprocessing algorithm is used first to improve the contrast and quality and reduce the noise of input brain MRI. Next, the statistical and texture features are retrieved from the preprocessed image to streamline the classification process. Consequently, the Rapid Sine Cosine Swarm Optimization (RSCSO) mechanism is used to decrease the dimensionality of features to speed up training and improve classification accuracy. Then, an Extreme Learning Machine (ELM) algorithm is employed to precisely forecast the photos of the healthy and tumor-affected tissues. Finally, an Auto Encoder-based segmentation process is used to accurately locate and crop the tumor-affected region from the aberrant images. During performance analysis, the proposed RSCSO-ELM mechanism's results are validated and compared using different measures and datasets.

Keywords

Brain Tumor, Computed Aided Diagnosis (CAD), Magnetic Resonance Imaging (MRI), Iterative Preprocessing, Statistical & Texture Features, Rapid Sine Cosine Swarm Optimization (RSCSO), Extreme Learning Machine (ELM), and Auto-Encoder.