Automatic Lung Segmentation in HRCT Images
This paper proposes a fully automatic algorithm for accurately segmenting lung regions in High-Resolution Computed Tomography (HRCT) images. A region based intensity characterization allows the image to be considered as consisting primarily of three regions: the CT background, the lungs, and the thorax region surrounding the lungs. A novel method based on flood fill algorithm is used to effectively identify the surrounding region. This step facilitates the use of another fast method for the removal of the CT background using linear scans originating from border pixels. Connected components that represent parts of the trachea are removed by noting the separation of the mean and standard deviation of intensity values between the trachea and the lungs. The segmented lung images are further enhanced to restore the intensity values of the pixels on the bronchi and the lung boundary. The proposed technique is not only computational inexpensive, but also robust and accurate in detecting the lung boundary. This paper presents the complete framework including examples and experimental results.