Generalized method for computation of true thickness and x-ray intensity information in highly blurred sub-millimeter bone features in clinical CT images

Abstract

In clinical computed tomography (CT) images, cortical bone features with sub-millimeter (sub-mm) thickness are substantially blurred, such that their thickness is overestimated and their intensity appears underestimated. Therefore, any inquiry of the geometry or the density of such bones based on these images is severely error prone. We present a model-based method for estimating the true thickness and intensity magnitude of cortical and trabecular bone layers at localized regions of complex shell bones down to 0.25 mm. The method also computes the width of the corresponding point spread function. This approach is applicable on any CT image data, and does not rely on any scanner-specific parameter inputs beyond what is inherently available in the images themselves. The method applied on CT intensity profiles of custom phantoms mimicking shell-bones produced average cortical thickness errors of 0.07 ± 0.04 mm versus an average error of 0.47 ± 0.29 mm in the untreated cases (t(55) = 10.92, p ≪ 0.001)). Similarly, the average error of intensity magnitude estimates of the method were 22 ± 2.2 HU versus an error of 445 ± 137 HU in the untreated cases (t(55) = 26.48, p ≪ 0.001)). The method was also used to correct the CT intensity profiles from a cadaveric specimen of the craniofacial skeleton (CFS) in 15 different regions. There was excellent agreement between the corrections and µCT intensity profiles of the same regions used as a 'gold standard' measure. These results set the groundwork towards restoring cortical bone geometry and intensity information in entire image data sets. This information is essential for the generation of finite element models of the CFS that can accurately describe the biomechanical behavior of its complex thin bone structures.

Model-based PSF and MTF estimation and validation from skeletal clinical CT images

“…However, it is conceivable that the PSF shape assumption may be adapted to non-Gaussian shapes. As discussed in the original introduction of this method, 24 the only information required to effectively apply it are anatomical knowledge of the region for most ideal selection of thincortical features, such as the general range of HU intensity values expected in the cortical bone.…”

“…It has been widely adopted in literature that the blurring process in CT scanners can be approximated by convolution of the ideal nonblurred image with a Gaussian function. 17,[20][21][22][23]8,24,25,9,26 Furthermore, for simplicity it is assumed that the point-spread-function (PSF) is reasonably approximated to exhibit radial symmetry within the slice plane and shift-invariance in all dimensions. As such, the PSF of the imaging system can be modeled as a normalized Gaussian function given by…”

“…(6) using a nonlinear optimization solver is outlined in detail. 24 To achieve stable solutions, the variables are explicitly bounded within conservative, but anatomically relevant windows; for example, the Y variables representing peak intensities for cortical and trabecular bone may only fall within 800-1800 and 100-800 HU, respectively. Similarly, the thicknesses for craniofacial cortical bone are within 0.2-3 mm.…”

Experimental validation of finite element predicted bone strain in the human metatarsal

“…This limitation may be overcome by using algorithms similar to those suggested by Pakdel et al (223), which estimates local dimensions of sub-mm bone features whose images are degraded by the limited resolution of clinical CT scanners. Another limitation of the current study is that metatarsals were loaded at a fixed angle, in a single mode of loading (i.e., bending), at a relatively low magnitude.…”

Left–right differences in the proximal femur’s strength of post-menopausal women: a multicentric finite element study

“…We may try to interpret this finding through an example: if contralateral bones are quite symmetrical (both for geometry and density distribution) then their strength will be similar irrespective of the loading condition imposed; if on the contrary contralateral bones show notable differences in geometry and density distribution, then their strength in a given loading condition will likely be significantly different, while seeking for minimum strength among a range of plausible loads may discover different patterns of structural weakness, thus limiting differences in minimum strength. Deeper studies, trying to single out the often subtle effects of geometry and density combinations to structural weakness, possibly incorporating information on cortical/trabecular compartments [34, 35, 36], are needed to prove this speculation. Nonetheless we believe this observational finding further corroborates the use of a minimum strength approach.…”