Feature-Extraction Methods for Lung-Nodule Detection: A Comparative Deep Learning Study

Abstract

Feature extraction has become a prerequisite step in computer vision problems, its importance resides in extracting significant hidden features from data, to help machine learning algorithms reach higher performance. Feature extraction techniques were behind the breakthrough in deep learning era, by providing relevant features. Deep learning architectures have overcome the state of the art in many different computer vision fields. In this work we are going to discuss and compare the accuracy of various global feature extraction methods, using deep learning for lung nodule detection. The experimental results show that feature extraction with convolutional neural networks (CNNs) outperforms the other methods including restricted boltzmann machines (RBMs).