FULLY AUTOMATIC REMOVAL OF CHEST TUBE FIGURES FROM POSTERO-ANTERIOR CHEST RADIOGRAPHS

摘要

The presence of artificial objects in radiographic images is common. For example, 33% of chest radiographs contain catheters. Anomaly detection algorithms used to monitor disease progression should not be confused by artificial objects such as catheters, chest tubes, pacemakers or even cloths that might be present in chest radiographs. Hence, the detection and the removal of artificial objects via a preprocessing module are very useful for Computer Aided Diagnosis (CAD) research. In this paper, we propose a Convolutional Neural Network (CNN) architecture that works as a trainable filter that removes simulated chest tube figures from chest radiographs.