Reproducibility of a semi-automatic method for 6-point vertebral morphometry in a multi-centre trial

Purpose: To evaluate the reproducibility of a semi-automated system for vertebral morphometry (MorphoXpress) in a large multi-centre trial. Materials and methods: The study involved 132 clinicians (no radiologist) with different levels of experience across 20 osteo-centres in Italy. All have received training in using MorphoXpress. An expert radiologist was also involved providing data used as standard of reference. The test image originate from normal clinical activity and represent a variety of normal, under and over exposed films, indicating both normal anatomy and vertebral deformities. The image was represented twice to the clinicians in a random order. Using the software, the clinicians initially marked the midpoints of the upper and lower vertebrae to include as many of the vertebrae (T5-L4) as practical within each given image. MorphoXpress performs the localisation of all morphometric points based on statistical model-based vision system. Intra-operator as well inter-operator measurement of agreement was calculated using the coefficient of variation and the mean and standard deviation of the difference of two measurements to check their agreement. Results: The overall intra-operator mean differences in vertebral heights is 1.61 ± 4.27% (1 S.D.). The overall intra-operator coefficient of variation is 3.95%. The overall inter-operator mean differences in vertebral heights is 2.93 ± 5.38% (1 S.D.). The overall inter-operator coefficient of variation is 6.89%. Conclusions: The technology tested here can facilitate reproducible quantitative morphometry suitable for large studies of vertebral deformities.