This study investigates the feasibility of using Xsight Diaphragm Tracking (XDT) for liver tumors located near or in the diaphragm during CyberKnife Synchrony Treatment. Eight patients (46 fractions) with liver tumors located near or in the diaphragm were reviewed retrospectively. Prior to motion data analysis, baselines were flattened to remove the portions with significant changes and improve accuracy. To reduce the uncertainty about sudden irregular breathing, respiratory data were calculated as a rolling average. The overall tracking accuracy based on the patient-specific respiratory curve was evaluated using E2E testing with CIRS (18023-A) dynamic phantom. Three main trajectories were observed in this work: Linearity (1/8), Linear radial type (6/8) and Hysteresis (1/1). The mean amplitude was 8.56±4.54 mm, 2.77±2.83 mm, and 4.23±3.92 mm for S-I, L-R, and A-P components. The linear trajectory patient had a more concentrated amplitude distribution. The baseline shifts were 5.88 mm (S-I), 2.53 mm (L-R), and 3.48 mm (A-P). Except for patient 2, all standard deviations of the center phase shift were less than 1 mm. The values of XDT correlation and prediction errors were 1.38±0.65 mm vs. 0.65±0.16 mm (S-I), 1.28±0.48 mm vs. 0.34±0.10 mm (L-R), and 0.96±0.32 mm vs. 0.22±0.072 mm (A-P), respectively. The strong positive correlations were amplitude vs. prediction error, SD of center phase vs. prediction error, and SD of center phase vs. amplitude. The median patient curve-based targeting accuracy was less than 1mm. Additionally, the mean target coverage for all patients with a 3 mm margin was 98.03±1.54%. This study proved that the diaphragm could be used as a tracking surrogate for liver tumors located in or near the diaphragm instead of placing golden fiducial markers. A reduction in motion amplitude and respiration training were necessary during liver SABR treatment, along with respiration control and evaluation.