Calculation of Cross-Sectional Geometry of Bone from CT Images with Application in Postmenopausal Women

Publisher Website
Google Scholar
Abstract
Objective The objectives of this study were to develop a software program that calculates bone cross-sectional area (CSA) and moments of inertia from digital images, test the computations for accuracy, and implement the software in a study of longitudinal changes in bone properties in a large group of postmenopausal women. Materials and Methods A program called SLICE was adapted to use computed tomography (CT) bone images as input. The main outcome variables of the resulting program, SLICE_CT, were bone CSA, areal moments of inertia, and mass-weighted moments of inertia. SLICE_CT was tested for accuracy using representative arrays of CT numbers that approximate cross-sections of cortical bone. Then the technique was applied to images from a 3 year longitudinal study of the radius in 86 postmenopausal women who underwent quantitative CT of the dominant radius at 30% the distance from the wrist to the elbow. The mean age of the subjects at entry into the study was 57 years. Images at baseline and at 3 years were analyzed, and the null hypothesis of no change in cross-sectional geometry was tested by a paired difference two-tailed t test. Results The error of SLICE_CT was <1% with proper segmentation, using half-maximal height as the threshold. In the longitudinal study of postmenopausal women, the mean change of CSA of the radius was – 1.4% (SD = 3.0%; p < 0.001). However, there was a compensatory 2.4% increase in the cross-sectional second polar moment of inertia (SD = 3.7%; p < 0.001). Conclusion SLICE_CT is an accurate method with which to determine cross-sectional geometry of digital images and can be used to detect changes in cross-sectional geometry of the radius with age in postmenopausal women. The results indicate that the radius remodels to compensate for bone loss and decreasing material properties of bone associated with aging.

This publication has 0 references indexed in Scilit: