Lung field segmenting in dual-energy subtraction chest X-ray images.

The purpose of this study was to develop and test a method to delineate lung field boundaries in dual-energy chest x-ray images. The segmenting method uses soft-tissue images and spatial frequency-dependent, background-subtracted images. Large-scale chest anatomy features are located and used to select the lung apices, the lateral lung boundaries, and the lung-mediastinum and lung-diaphragm boundaries. Extraneous parts of the contours are removed and they are joined to form complete lung boundaries. The reliability measure uses a statistical shape model to estimate the probability of occurrence of a contour. The method was experimentally tested with 30 human subject images. It has higher accuracy and specificity and a sensitivity parameter equal to the best previously reported method. The reliability measure is able to detect contours with unusual lung outlines or errors in the processing. The method exploits the characteristics of dual-energy subtraction images to improve lung field segmenting performance.