Using Heuristics for the Lung Fields Segmentation in Chest Radiographs

Abstract

The cancerous attaches, mainly the lung cancers, make a serious medical problem all over the world. The early diagnoses based on chest radiographs could notably lower its mortality. The efficiency of the computers gives the possibility to facilitate the work of the radiologists by a CADsystem. But first the region of the interest, i.e. the lung fields should be determined. The lung segmentation in our sense differs from the trends accounted in literature (where the area hidden by the heart is ignored), because the left border of the left lung is located beneath the heart. In this paper we describe a method based mainly on heuristics and rules which can be used to find the contours of the lung. The algorithm is divided to five main steps: (1) finding some parameters of the lungs without long processing, (2) determining the usual lung contours, (3) finding the mediastinum, (4) finding the lower border of the left lung and (5) applying a model to achieve better results, to make some refinement.