Brain Tumors MRI Efficient Feature Dimensionality Reduction and Classification Using IELM and OSVM

Author (s) : Parasuraman Kumar and B. Vijayakumar

---

Abstract :
Magnetic resonance imaging approach differentiates and classifies the neural architecture of human brain. MRI methodology has numerous imaging modalities, which perform the scanning and capturing of the intrinsic structure of human brain. In this research work, the focus is on noise removal approach, extraction of gray‐level cooccurrence matrix (GLCM) features, DWT‐based brain tumor region growing segmentation for reducing the complexity and boosting the performance. Then this is follow by enhanced morphological filtering that eliminates the noise, which can be introduced after segmentation. Dimensionality reduction is performing utilizing the bacterial foraging optimization (BFO) algorithm and in addition the number of feature space is also decreased. Automated brain tumor stage classification is carried out with Improved Extreme Learning Machine (IELM) and Online Support Vector Machine (OSVM). This phase categorizes the brain images into tumor and non‐tumors with hybrid ensemble based classifier. Experiments have shown that the technique was more reliable for initialization, quicker and accurate.

Keywords : IELM, OSVM, DWT, BFO and GLCM.

Conference Date : 26-04-2019

Pages : 21-29

Paper ID : 8510046