In X-ray CT, projections are sometimes truncated due to size mismatches between small detectors and/or large objects, which introduce truncation artifacts in reconstructed images. In this work, we aim at addressing the truncation problem using the idea of “big data”-aided analysis/synthesis. For a specific cross-section position of a patient, e.g., pelvis in this paper, we collect hundreds CT images to build a database. We propose the “image-sum” as the feature of each image in the database. With this feature, we can find the most relevant image to the truncated projection from the database. Finally, the forward projections of the selected image are used to extend truncated projections. The experimental results show that the proposed method can reduce the truncation artifacts successfully.