A tumor recognition method for 3D CT image data volume,which can accurately identify the tumor in CT images,is developed in this project.This method organizes the PET-CT images of the tumor patient by the mutual information registration theory and implements the adaptive threshold classification algorithm to classify the region of interest(ROI)in PET images.Then,the 3 dimension texture information analysis method and edge segmentation processes are applied to recognize the accurate position of the tumor and extract its edge in 3D CT data volume(axial and sagittal CT image).Finally,the tumor edge of sagittal image can be used to refme the edge of the same tumor in axial image.The broken or inaccurate tumor edges can be fixed and refine by the texture energy analysis of the 3D CT data volume.