3D　shape　reconstruction　of　the　hepatic　tumor　from　its　2D　cross-sections　improves　the　surgeon＇s　knowledge　of　tumor　anatomy　and　makes　even　more　complicated　tumor　surgery　safe.　Image　segmentation　and　3D　reconstruction　are　two　key　technologies　for　tumor　visualization.　Being　aimed　at　the　difficult　problem　about　how　to　effectively　segment　hepatic　tumor　from　CT　images,　the　graph-theory　segmentation　algorithm　is　first　attempted　to　automatically　extract　tumor.　In　addition,　a　novel　3D　reconstruction　method,　which　effectively　combines　New　Marching　Cubes　algorithm　with　graph-theory　auto-segmentation　algorithm,　is　proposed　and　employed　to　visualize　tumor　surface.　The　proposed　method　has　been　evaluated　via　CT　image　series　of　6　patients　suffering　from　liver　cancer.　The　results　demonstrate　that　segmentation　accuracy　is　very　high　where　the　average　minimum　Euclidean　distance　and　area　overlap　measurement　are　0.101　and　0.965,　3D　reconstruction　is　fast　and　the　tumor　model　is　also　relatively　satisfactory,　and　it　can　be　employed　to　aid　clinical　practice　as　an　alternative　tool.