Lung Nodule Classification Using Deep Features in Ct Images

This paper suggest a novel arrangement technique for the four varieties of lung nodules, i.e., well- circumscribed, vascularized, juxta-pleural, and pleural-tail, in minute dose computed tomography scans. The proposed technique is built on contextual analysis by merging the lung nodule and adjacent anatomical structures, and has three main stages: an adaptive patch-based division is used to build concentric multilevel partition; then, a new feature set is planned to incorporate intensity, texture, and gradient information for image patch feature description, and then a contextual latent semantic analysis-based classifier is designed to compute the probabilistic estimations for the related images. Our proposed method was estimated on a publicly existing dataset and obviously demonstrated promising classification performance.