Automated liver segmentation using multislice CT images

Abstract

AbstractWith practical applications of multislice CT, high‐speed imaging has drastically become possible in recent years as compared with the past and imaging the liver twice during the time a patient holds his breath once has become possible in liver CT imaging using contrast media. Thus, three‐dimensional images of two time phases differing in the circulatory state of a contrast medium are obtainable without positional discrepancies. The authors propose an automated liver area extracting (segmenting) method using three‐dimensional CT images of two time phases of a contrast medium for extracting the shape of the liver. This method first constructs a two‐dimensional characteristic (feature) space having the CT values of time phases on each axis from the three‐dimensional images of a two time phase portion imaged immediately after and 10‐plus seconds after the infusion of a contrast medium. The method extracts areas showing changes in the contrast medium corresponding to the liver in this space and generates images with the liver region emphasized using these. The method then determines the liver region by performing opening and closing procedures on the liver area‐emphasized images generated. Finally, it determines the liver region by a three‐dimensional component connecting procedure using geometric shape characteristics of the blood vessels, tumors, and so on. The efficacy of using images of two time phases simultaneously has been verified by applying the proposed method to nine cases. © 2003 Wiley Periodicals, Inc. Syst Comp Jpn, 34(9): 71–82, 2003; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/scj.10210