Lung Nodule Image Classification Based on Ensemble Machine Learning

This study investigates the problem of lung nodule classification in low dose CT images. A novel method is proposed based on ensemble machine learning. The method includes three stages. First, diverse feature extraction and representation along with the classification algorithm are selected, and initial candidate classifiers are trained. Second, the ensemble classifier set is obtained based on the well defined fitness function, which is used to measure the comprehensive performance of the selected classifier set. At last, the final classifier is build in the light of aggregation on ensemble classifier set. Extensive experiments are evaluated and careful reports demonstrate the effectiveness of our proposed ensemble based classification approach.