Comparing Methods for Calculating Cross-Sectional Properties From High-Resolution CT Scans in Dry Bone: The Case of the Non-Adult Femoral Midneck

Objective

This paper explores the various protocols for deriving endosteal and periosteal contours of the femoral midneck cross-section in non-adult individuals using EPmacroJ.

Materials and Methods

We analyzed femoral midneck cross-sectional contours in a total sample of 55 femora belonging to medieval individuals ranging in age between 2 and 20 years. Femoral midneck cross-sections were taken on high-resolution CT images and analyzed by EPJMacro in FIJI. Cross-sectional properties were obtained from periosteal and endosteal contours derived by manual, automatic, spline, and ellipse techniques. Agreement between the manual and the other techniques was evaluated using percent prediction error (%PE) and following the Bland–Altman approach.

Results

All the evaluated techniques used for deriving the periosteal and endosteal contours at the femoral midneck cross-section in non-adult individuals have proven to be interchangeable with manual processing, except for the spline technique for the endosteal contour. The narrowest limits of agreement were obtained for the total area, moderate limits of agreement were found for the cortical and medullary areas, and the widest limits of agreement were observed in the section moduli.

Discussion

The automatic processing of the periosteal and endosteal contours shows a high level of agreement with the manual processing of both contours. Semi-automatic approaches (spline and ellipse) provide less agreement, especially in section moduli. This suggests that inaccuracies in deriving the endosteal contour significantly affect section moduli. The best approach for extracting the geometric properties of femoral midneck cross-sections is a combination of automatic and manual methods, although automatic methods alone are also effective.