Medical Image Classification based on Ensemble Machine learning Methods

Abstract

Lung cancer was classified as the most danguresous type of cancer due to it’s high rate of deaths. Lung nodules can either be benign so noncancerous or can be can- cerous and so cause lung cancer. Detecting these nodules in early stages highly increases the treatment probability and so saves human precious lives. From the beginning of the new century, artificial intelligence and machine learning tech- niques began to have a hand in our daily life and interfere in almost every thing to make it much easier and more efficient. Those techniques made a great impact in the medical field no one can ignore it. Here in this thesis we used machine learning technique to help in detecting and localizing these lung nodules with fast and accurate results. Random forest is one of the most important machine learn- ing techniques that proved its efficiency in many fields. Through out the different chapters of this thesis we aimed to fulfill our goal which is our title “Medical Images Classification Based on Ensemble Machine Learning Methods” and cover wide aspects of interest. Firstly, a comprehensive survey and analysis for lung nodule detection and classification using RF was proposed. The most popular used datasets have been listed, commonly used features were presented, some of the most important preprocessing steps were listed. The recent models and the remarkable studies with their achieved results were reported. From that review chapter we concluded that selecting suitable features and choosing accurate learner for a specific classification task can vary widely depending on different criteria such as the addressed problem, used dataset and the desired output. Then we took RF