A visual saliency-based method for automatic lung regions extraction in chest radiographs

Extracting lung regions accurately from a chest X-ray is an important procedure in computer-aided lung disease diagnosis. The shape and size of lungs may hold clues to serious diseases such as pneumothorax, pneumoconiosis and even emphysema. However, the precise extraction of lungs from a X-ray is still very difficult at the moment. In this paper, we propose a novel method of detecting the lung regions in chest radiographs. It is based on the observation that the lung fields in X-ray images well stand out against the background which makes them salient regions. According to our method, a X-ray image of lung is firstly segmented into several small sub-regions through graph-based segmentation. Then we detect the salient value of each sub-region using a global contrast function. The lung region can be estimated based on the salient values of each sub-region. Finally, cubic spline interpolation is used to obtain smoother boundaries by refining the results. In the experiment, we built a Lung Region Location model including 147 randomly selected chest X-ray images from the JSRT dataset and used the remaining 100 images in it to test our method. The results demonstrate that our method achieved state-of-the-art performance.