A Framework for Automatic Segmentation of Lung Nodules from Low Dose Chest CT Scans

To accurately separate each pulmonary nodule from its background in a low dose computer tomography (LDCT) chest image, two new adaptive probability models of the visual appearance of small 2D and large 3D pulmonary nodules are jointly used to control the evolution of the de-formable model. The appearance prior is modeled with a translation and rotation invariant Markov-Gibbs random field of voxel intensities with pairwise interaction. The model is analytically identified from a set of training nodule images with normalized intensity ranges. Both the nodules and their background in each current multi-modal chest image are also modeled with a linear combination of discrete Gaussians that closely approximate the empirical marginal probability distribution of voxel intensities. Experiments with real LDCT chest images confirm the high accuracy of the proposed approach