Assessment of bone fracture risk is the first step in the prevention of traumatic events. In several previous study the use of bone mineral density and bone volume fraction was suggested for the identification of the failure zone, nonetheless the limits of this approach were also investigated, underling the need of other information to fully describe the failure event. In the present study, a comparison between fracture and non-fracture zones of trabecular bone is proposed with the aim of analyze the local structural differences attempting to identify the morphometrical parameters who best can describe the trabecular fracture zone. Eighteen trabecular specimens were extracted from the lower limb of two donors without skeletal disorders. All the specimens were scanned by means of a micro-CT and mechanically tested. After the mechanical compression every specimen was scanned again obtaining for every specimen two datasets: pre- and post-failure. An automatic registration scheme, comprising of a three-dimensional automatic registration method to define the differences between the two datasets, and the application of a criterion for defining "broken" or "unbroken" trabeculae, was applied for the identification of the full 3D fracture zone. The morphometrical analysis of fracture and non-fracture zone was performed by the study of several morphometrical parameters, such as bone volume fraction, off-axis angle, structural model index, connectivity density, etc. The results of the two different structures were compared by means of a Wilcoxon non-parametric test. Ten out of 12 morphometrical parameters were found statistically significantly different between fracture and non-fracture zones, underlining the strong structural difference between the two areas. Nonetheless, only three of them have shown differences superior to 30%, with a reduce overlapping of their distributions: off-axis angle, structural model index and connectivity density. On the other hand, bone volume fraction showed a smaller, even if significant, difference with great overlap of the distributions, in agreement with the limits already pointed out in the literature.