Vertebra Location-Based Multi-Regional Segmentation of Liver CT Images

Liver segmentation has always been difficult in CAD (Computer-aided diagnosis). Especially, in different layers of liver CT images there may be two or more regions, which add a more arduous problem to liver segmentation. To solve this problem, a new automatic segmentation method for multi-regional liver CT image is presented. That is, first using a preprocessing way based on multi-threshold, morphological methods as well as position information to segment the vertebra and then according to the information of a fixed vertebra location and liver, selecting several seeds for the rough segmentation of liver, meanwhile, using voting mechanism to solve the problem for the single seed falling into the trap. Finally, by using morphology to amend the rough segmentation result we could eliminate the problem of over-segmentation. Actual segmentation experiments of abdominal CT image verify the effectiveness of the algorithm proposed in this paper.