Lung Field Detection on Chest Radiographs

M. Monteiro, A.M. Mendonça, and A. Campilho (Portugal)


Chest radiographs, image segmentation, computer-aided diagnosis, image analysis


The purpose of the research herein presented is the automatic detection of the lung boundaries in posterior anterior digital chest radiographs. The precise location of the two lungs is important in a computer-aided diagnosis system as it allows the reduction of the region under analysis, decreasing the computing time and facilitating data compression. Furthermore it allows the delimitation of the search area easing the selective tuning of the abnormalities detection algorithms. We adopted an edge based image segmentation technique, by iteratively determining the costal-mediastinal and lower-upper lung boundaries, followed by local spline fitting to each one of these anatomical boundaries. The final results are two closed contours, delineating the two lung fields. A good accuracy of the method was obtained in a test set of forty seven chest images.

Important Links:

Go Back