User Variance in Defining the Trabecular‐Cortical Interface Using MicroCT Images

Abstract

Micro‐computed tomography (µCT) imaging is an effective tool for quantifying bone microarchitecture. µCT images allow trabecular and cortical bone to be separated, a task critical to understanding bone mechanics. Automated algorithms are often used to facilitate this process, but are typically evaluated by comparing algorithm results to those of a single person. Such analyses do not take into account the possibility of variability among observers in defining this boundary.This study quantified the amount of inter‐observer variation in defining the cortical‐trabecular bone boundary. Ten µCT images of human femora and primate mandibles were sent to 19 researchers with interests in bone biology. Participants were asked to trace the cortical‐trabecular boundary on each image as well as to complete a questionnaire about education, training, and specific areas of interest in bone biology. For each image, boundary lines were converted to semilandmarks for use in geometric morphometric analyses. Shape variation was summarized using principle components analysis and variation among users was quantified.All of the µCT images exhibited variance in where the participants defined the cortical/trabecular interface. Often the variation was small (<10 voxels), but in some images there was considerable variation in boundary location. Analyses did not reveal significant differences (p>0.05) in boundary contours based on different levels of education or areas of expertise. The results do, however, indicate variability among bone biologists in defining the trabecular‐cortical boundary and question the validity of algorithms evaluated against a single user.