Computer-Aided Lung Parenchyma Segmentation Using Supervised Learning

Abstract

Advances in the fields of image processing and information technology have led to the use of computers for the diagnosis of diseases. This has led to the emergence of Computer-Aided Diagnosis (CAD) systems for disease diagnosis. This research work focuses on improving the performance of CAD systems that use Computed Tomography (CT) of the chest for diagnosis of lung disorders. This improvement has been achieved by developing techniques for determining the significance of image features for discrimination among different diseases of the lung and a technique for segmentation of lung parenchyma in chest CT irrespective of the presence or absence of peripherally placed Pathology Bearing Regions (PBRs). Another major challenge in CAD of lung disorders based on analysis of chest CTs is accurate segmentation of lungs especially in the presence of peripherally placed PBRs. In this research work, a segmentation algorithm has been developed to extract the complete lung parenchyma even in the presence of severe peripherally placed PBR in chest CT. The proposed system has been found to improve the diagnostic performance of CAD systems for diagnosis of lung disorders based on analysis of chest CT slices. This would aid the physicians to perform better diagnosis, which would result in choosing the appropriate treatment strategy, thereby reducing the mortality rate.