BACKGROUND: There is no non-invasive in vivo method to assess intervertebral kinematics. Current kinematics models are based on in vitro bone reconstructions from computed tomography (CT)-scan imaging, fluoroscopy and MRIs, which are either expensive or deleterious for human tissues. Musculoskeletal ultrasound is an accessible, easy to use and cost-effective device that allows high-resolution, real-time imaging of bone structure. OBJECTIVE: The aim of this preliminary study was to compare the concordance of 3D bone modeling of lumbar vertebrae between CT-scan and ultrasound imaging and to study the intra and inter-reliability of distances measured on 3D ultrasound bone models. METHODS: CT-scan, ultrasound, and in situ data of five lumbar vertebrae from the same human specimen were used. All vertebrae were scanned by tomography and a new musculoskeletal ultrasound procedure. Then, 3D bone modeling was created from both CT-scan and ultrasound image data set. Distances between anatomical bones landmarks were measured on the 3D models and compared to in situ measurements. We observed that all distances were included within the limit of agreement between the three methods of measurements (3D CT-scan, 3D MSU and in situ) with a good intra- and inter-reliability of 3D ultrasound measurements of 0.97 and 0.82, respectively. Based on the mean of mean differences between methods, we observed for all distances, ultrasound overestimated distances of 0.44 ± 0.63 mm compared to CT-scan, ultrasound underestimated distances of 0.39 ± 0.48 mm compared to in situ measurements and CT-scan underestimated distances of 0.93 ± 0.55 mm compared to in situ measurements. CONCLUSIONS: Three-dimensional modeling from ultrasound imaging is similar in comparison to 3D bone modeling from CT-scan imaging with a good intra and inter reliability.