Automatic Detection of Abnormalities in Magnetic Resonance Imaging

Abstract

The thesis has six chapters organized as follows: Chapter 1 Introduction: introduces the proposed study for detecting Multiple Sclerosis (MS) using Ensemble Machine Learning, in addition, presents the research scope, the objective of the study, and the contribution to improve the performance of MS automatic early detection.

Chapter 2 Background: discusses the biological and theoretical background of topics involved. These background topics include the MS disease definition, Magnetic Resonance Imaging (MRI) Sequences, Machine Learning methods (Ensemble Support Vector Machines and Ensemble Decision Tree), Feature Extraction Techniques (2D-DWT method, Textural Features).

Chapter 3 Literature Review: shows a survey that covers all studies that presented the automatic segmentation using machine learning methods.

Chapter 4 Methodology: presents the proposed model's details that combine the feature-based with ensemble machine learning methods to create a fully automated model used to detect MS in MR brain image. Additionally, this chapter explains in detail the flow chart of the model (Preprocessing, Processing, Postprocessing) and the function of each step.

Chapter 5 Results & Discussion: discusses the two methods' results and compares results between two models (ESVM, EDT). Additionally, this chapter discusses the challenges and problems we face in the study and how to solve it optimally.

Chapter 6 Conclusion & Future Work: gives a conclusion for the thesis work, and potential directions for future work.

Keywords: Magnetic Resonance Imaging, Multiple Sclerosis, Automatic segmentation, Ensemble support vector machine, Ensemble Decision Tree, Lesions, RUSBOOST , Cost-Sensitive Learning.