Automatic 3D segmentation of the liver from computed tomography images, a discrete deformable model approach

Automatic segmentation of the liver has the potential to assist in the diagnosis of disease, preparation for organ transplantation, and possibly assist in treatment planning. This paper presents initial results from work that extends on previous two-dimensional (2D) segmentation methods by implementing full three-dimensional (3D) liver segmentation, using a self-reparameterising discrete deformable model. This method overcomes many of the weaknesses inherent in 2D segmentation techniques, such as the inability to automatically segment separate lobes of the liver in each image slice, and sensitivity to individual-slice noise. Results are presented showing volumetric and overlap comparison of twelve automatically segmented livers with their corresponding manually segmented livers, which were treated as the gold standard for this study