This study aims to quantify in vivo deformation of stented coronary vessel centerlines due to cardiac motion to understand the potential errors in fusing optical coherence tomography (OCT) with angiography. We first evaluated the static error and test the reproducibility of a vessel centerline reconstruction method derived from the stereoscopic theory in vitro and in vivo. Two phantom models mimicking coronary artery bifurcations were used for in vitro static conditions, and four coronary arteries (2.75 mm ± 0.14 mm) of two Yorkshire swine implanted with 3.0 mm × 17 mm bare metal stents were used for in vivo dynamic conditions. Our method depicted a strong linear correlation (R² = 0.91) between the reconstructed geometry and the actual geometry with the error of 2.3 mm ± 1.8 mm across various angles between paired images (50°-130°) in vitro. This method also showed higher accuracy in the stented segment's length and curvature compared to currently available methods (the root mean square error = 0.76 pixel vs. 1.3 pixel), and good reproducibility across various angles (50°-130°) and in two different cardiac cycles in vivo. The reconstructed vessel centerlines did not deform significantly over a cardiac cycle in vivo (error of length = 0.17 mm ± 0.16 mm, maximum curvature = 0.13 ± 0.09 with error = 0.07 ± 0.06 in the stented segment; four different cardiac phases, 6 ± 2 time-points). Despite small sample size, the results may support using the vessel centerline as a fusion path for non-ECG-gated intravascular images, such as OCT images, because cardiac motions introduce only a small error in the vessel centerline reconstruction.