Document Type


Publication Date



The analysis of X-ray imagery is the standard pre-screening approach for lung cancer. Unlike CT-scans, X-ray images only provide a 2D projection of the patient’s body. As a result occlusions, i.e. some body parts covering other areas of the body within this projected X-ray image, can make the analysis more difficult. For example, the ribs, a predominant feature within the X-ray image, can cover up cancerous nodules, making it difficult for the Computer Aided Diagnostic (CAD) systems or even a doctor to detect such nodules. Hence, this paper describes a methodology for reconstructing a patient-specific 3D model of the ribs and lungs based on a set of lateral and PA X-ray images, which allows the system to calculate simulated X-ray images of just the ribs. The simulated X-ray images can then be subtracted from the original PA X-ray image resulting in an image where most of the cross hatching pattern caused by the ribs is removed to improve on automated diagnostic processes.


© Christopher Koehler and Thomas Wischgoll licensed under Creative Commons License NC-ND.