Objective: To propose and validate a correction method for the respiratory phase shift of using different motion management systems in the 4-dimensional (4D) imaging and free-breathing gated radiotherapy. Materials and methods: Synchronized cone-beam CT (CBCT) and optical surface images (OSI) of 30 patients at two institutions were included. Reference diaphragm-signals were extracted from CBCT projections using Amsterdam-Shroud (AS) method. Principal Component Analysis (PCA) was used to reduce the dimensionality of OSI data. The phase shift between the AS and PCA signals was calculated using the cross-correlation function. The 20 patients in public 4D-Lung dataset were used for external validation. The relative phase shift averaged over the dataset was proposed as the correction factor. Based on two signals, 4DCBCT were reconstructed using Feldkamp-Davis-Kress (FDK) algorithm respectively. Results: Phase discrepancies of 396 f 106 ms were observed in the clinical patients. The OSI signals consistently lagged behind the diaphragm signals by 12 % f 3 % (clinical) and 12 % f 6 % (public) of a respiratory cycle. Using 12 % as a correction factor, the phase shift was reduced to 76 ms (by 81 %) on average for the 30 clinical patients, and the Root Mean Square Error was significantly decreased by 10-23 % for the phase-averaged 4DCBCT. For 20 public cases, the tumor centroid displacement was reduced by 1.5 f 0.9 mm, and the intersessional variation was 2 % f 1 % of a respiratory cycle for the 10 cases with multiple 4DCT. Conclusion: The significant phase inconsistency using different motion management techniques should be monitored and corrected before navigating 4D imaging and treatment.