<p>Objective: Robot-assisted ultrasound (US) system could potentially provide a non-radiative three-dimensional (3D) reconstruction for minimally invasive spine surgery. Despite promising studies on this technology, few researchers provide a comprehensive analysis with technical details. The performance is only showcased in one specific experimental setting without offering insight into robustness and generality of the approach by e.g. verifying performance in various testing scenarios. This limitation impedes the translation of this technology toward clinical practice. Therefore, this study provides a comprehensive assessment of the performance of robot-assisted US system. This study provides all essential technical details from experiments starting at the bench-top up to pre-clinical study. Methods: A hybrid control strategy was proposed to ensure continuous and smooth scanning, while a U-Net based US reconstruction framework was utilized to process the anatomic features automatically. Experiments were conducted on two synthetic spine phantoms and two ex-vivo cadavers. Results: The average deviation of scanning force was 2.84 ± 0.45 N on the synthetic phantom and 5.64 ± 1.10 N on the ex-vivo cadaver. The reconstruction yielded a mean 3D representation error of 1.28 ± 0.87 mm and 1.74 ± 0.89 mm for the synthetic and cadaveric experiments, respectively. Conclusion: The experiments indicated the proposed robot-assisted US system is feasible for relatively independent experiment settings, varying from laboratory experiments to pre-clinical studies. Significance: The developed system offers an improved understanding of the potential for this technology and paves the way for deploying the robotic US system towards in-vivo clinical study.</p>