Abstract: Computed tomography (CT) images of stems from Tamarix ramosissima Ledeb. were used to construct a three-dimensional anatomical model of tree rings, and growth parameters, such as the tree-ring growth rate (TRGR) and annual timber volume (ATV) were determined. The research results have reference value for forest growth assessment and forest carbon sink measurement. In this study, CT and flat two-dimensional cross-sectional scan images of Tamarix ramosissima stem were obtained by CT and flat scanning, respectively. U-net semantic segmentation model of CT images was obtained using Python deep learning methods. Combining the U-net semantic segmentation model constructed in previous research for flat scan tree-ring images, early wood regions from two types of image sources were obtained, respectively. After images disc scanning and semantic segmentation, tree-ring indexes such as the circumference, basal area increment (BAI), TRGR, and ATV index were obtained by using GIS rectifying, editing and measurement tools. Based on CAD images by using CT images extract in spaced order, three-dimensional anatomical diagrams of Tamarix ramosissima tree-ring were reconstructed to characterize the tree rings structure by using Sketchup 3D software. Paired-sample t-tests, Pearson correlation coefficients, and Bland-Altman quantitative consistency analyses were used to evaluate the consistency of the growth-ring measurements obtained from CT and flat scanning of Tamarix ramosissima, and to verify the accuracy of the tree-ring parameters extracted from the CT scans. The study results indicated that TRGR indexes measured by CT and flat scanning were found to be consistent. The P-values of the paired-sample t-tests for comparing the quantitative means of the TRGR sequences were all greater than 0.05, indicating that there were no significant differences between the measurement results obtained from the two image sources. The Pearson correlation coefficient between the two sets of results was 0.9984 (P<0.0001), indicating a highly significant correlation. The Bland-Altman quantitative consistency analysis showed that 95.625% of the sample differences between the two measurement approaches were within the limits of consistency, indicating good consistency between the results obtained by the two methods. Thus, the growth parameters of Tamarix ramosissima tree-ring based on CT images and measured by GIS were demonstrated to be accurate. This method can be applied to the measurement of tree-ring growth parameters of different kinds of trees in agriculture and forestry. These results provide a reference for the development of non-destructive tree-ring measurement technology and the estimation of ATV in the future.