A modified bit-plane based method for lung region extraction from 3D chest CT images

Abstract

In this paper, we present an enhancement on bit-planes based method in the phase of extracting lung region from 3D chest images which is considered a key challenge in computer aided diagnosis systems designed for early detection of lung cancer. The chest 3D-CT images are used without any preprocessing in order to keep the entire information intact within the raw data. The results show that Bit-plane1 and Bit-plane2 can be considered together as needed and sufficient information to the extraction process of lung region even if other small objects having common pixels with lung walls remain within the extracted objects. Our method is implemented and tested on a database of normal cases composed of more than 1000 CT intensity images, and has successfully extracted lung regions in almost all of the cases.