On the assessment and use in modeling of the mechanical properties of vertebral trabecular bone calculated from computed tomography data

Osteoporosis is characterized by a decrease in the quality and volumetric bone mineral density (vBMD). vBMD can be determined based on Hounsfield units (HU) calculated from quantitative computed tomography (CT) data. CT data typically contains several series of images obtained using different convolution kernels. This article examines the influence of CT convolution kernels on the results of assessing the mechanical properties and stress-strain state of cancellous bone in the lumbar vertebrae. To this end, two anonymised series of CT images of the lumbosacral spine were analysed. For each series of images, two sets of mechanical properties of the vertebral bodies cancellous bone were determined, and two corresponding biomechanical models of the lumbosacral spine segment were constructed, together with a fixation system at the L4–L5 vertebral level. The stress-strain state of both models under typical static loads was simulated using the finite element method. The largest absolute difference in the corresponding HU was 52,2 for the L5 vertebra, and in the elasticity modules – 47 MPa. Modelling revealed that differences in maximum equivalent stresses in the cancellous bone of vertebral bodies amounted to 35 %. It has been shown that when analysing the results of modelling, a problem may arise related to the correctness of assessing the risk of damage to cancellous bone based on a comparison of the stresses arising in the bone and its strength characteristics. The results of the study emphasise the importance of taking into account CT image processing parameters when performing biomechanical analysis, especially in the context of planning orthopaedic surgery and predicting implant durability. The work also points to the need to standardise non-invasive methods for assessing the mechanical properties of bone tissue in order to improve the accuracy and reliability of patient-specific modelling.