Automatic vertebra tracking through dynamic fluoroscopic sequence by smooth derivative template matching

Diagnosis of the underlying causes of widespread spinal pathologies such as back pain and whiplash remains problematic. Many studies suggest that segmental instability may occur and that the study of the intervertebral kinematics can be a valuable, objective method to assess spinal segment functionality. Direct measurement of the intervertebral kinematics results very invasive and unpractical; as alternative analysis of dynamic videofluoroscopic can provide intervertebral kinematic data of lumbar and cervical spinal tracts during unconstrained patient motion, with an acceptable low X-ray dose. Estimation of the kinematics relies on accurate recognition of vertebra positions and rotations on each radiological frame; this can be achieved identifying specific feature points or landmarks, but manual selection results tedious and imprecise. The aim of this work is to present an improved procedure and automatic identification of vertebra motion. By opportunely processing the radiological sequences by using smoothed derivative operators the main vertebral body outlines results enhanced; thus, procedures of template matching for vertebra location become more accurate. Furthermore, data interpolation provided sub-pixel accuracy. Kinematic data, obtained by processing real sagittal fluoroscopic sequences of the lumbar spine, were tested against results of previous studies obtained by manual identification and other methods. Time-evolution of intervertebral kinematic parameters resulted less variable than the other methods; root mean square differences and standard deviations were computed. Vertebra trajectories were interpolated by smoothing cubic spline and instantaneous speed and acceleration were computed. Vertebra speed and acceleration resulted more stable, smooth and in accordance with the actual motion preformed by patients.