Automated identification of 3D lung CT image orientation

Abstract

Computerized tomography (CT) is one of the major high-resolution imaging modalities. Itis especially useful in the early detection of lung abnormalities. However, CT images are sometimes stored with names that do not always indicate the order of 3D images. Using “SliceLocation” and “InstanceNumber” in the CT header, CT image slices can be arranged into 3D form. However, the orientation of 3D images may still be reversed. The objective of the proposed method is to automatically determine the orientation of ordered 3D lung CT images by identifying the position of the trachea. The two features consisting of area difference and mean of roundness along several slices are extracted. The full dataset of LIDC-IDRI containing 10,010 3D lung CT images was evaluated to measure the performance of the proposed method. This proposed method achieves 99.97% accuracy. The proposed procedure can be very useful for development of computer aided detection or diagnosis of 3D lung CT images

Keywords

• automated identification
• cad
• data preparation
• 3d lung ct image orientation