Learning with distribution of optimized features for recognizing common CT imaging signs of lung diseases

• Type of Material: Article
• Authors: Ling Ma, Emory University; Xiabi Liu, Beijing Institute of Technology; Baowei Fei, Emory University
• Language: English
• Date: 2017-01-21
• Publisher: IOP Publishing
• Publication Version: Author Accepted Manuscript (After Peer Review)
• Copyright Statement: © 2016 Institute of Physics and Engineering in Medicine.
• Final Published Version (URL): http://dx.doi.org/10.1088/1361-6560/62/2/612
• Title of Journal or Parent Work: Physics in Medicine and Biology
• ISSN: 0031-9155
• Volume: 62
• Issue: 2
• Start Page: 612
• End Page: 632
• Grant/Funding Information: The work was also supported in part by the Georgia Research Alliance (GRA) Distinguished Cancer Scientist Award to BF.; This research was supported in part by U.S. National Institutes of Health (NIH) grants (CA176684 and CA156775).; XL was partially supported by National Natural Science Foundation of China (Grant no. 60973059, 81171407) and the Program for New Century Excellent Talents in Universities of China (Grant No. NCET-10-0044).; LM was partially supported by International Graduate Exchange Program of Beijing Institute of Technology.
• Keywords: TEXTURE CLASSIFICATION; GROUND-GLASS; CANCER STATISTICS; computed tomography (CT); SOLITARY PULMONARY NODULE; Science & Technology; Life Sciences & Biomedicine; Engineering; PERFORMANCE; image classification; Technology; DIFFERENTIATION; feature extraction; Engineering, Biomedical; Radiology, Nuclear Medicine & Medical Imaging; HIGH-RESOLUTION CT; EMPHYSEMA; AUTOMATED CLASSIFICATION; lung disease; COMPUTER-AIDED DIAGNOSIS
• Research Categories: Health Sciences, Radiology; Engineering, Biomedical

Abstract

Common CT imaging signs of lung diseases (CISLs) are defined as the imaging signs that frequently appear in lung CT images from patients. CISLs play important roles in the diagnosis of lung diseases. This paper proposes a novel learning method, namely learning with distribution of optimized feature (DOF), to effectively recognize the characteristics of CISLs. We improve the classification performance by learning the optimized features under different distributions. Specifically, we adopt the minimum spanning tree algorithm to capture the relationship between features and discriminant ability of features for selecting the most important features. To overcome the problem of various distributions in one CISL, we propose a hierarchical learning method. First, we use an unsupervised learning method to cluster samples into groups based on their distribution. Second, in each group, we use a supervised learning method to train a model based on their categories of CISLs. Finally, we obtain multiple classification decisions from multiple trained models and use majority voting to achieve the final decision. The proposed approach has been implemented on a set of 511 samples captured from human lung CT images and achieves a classification accuracy of 91.96%. The proposed DOF method is effective and can provide a useful tool for computer-aided diagnosis of lung diseases on CT images.