Replicating Expert Bone Measurements Automatically: A Case Study with Femoral Allografts

Abstract

Precise morphological assessment of bone allografts is essential for successful graft-recipient matching in tissue banking. Manual measurement protocols, while standard, are labor-intensive, time-consuming, and prone to inter- and intra-observer variability. To overcome these limitations, this work presents an automated system integrated into the BeST-Graft Viewer for performing femoral allograft measurements. The proposed method replicates expert measurement strategies by defining anatomical search zones and reference planes directly within the 3D model of the femur. Reference points are automatically detected based on geometric conditions, and the corresponding anatomical parameters are computed using predefined formulas. The system was validated against manual measurements performed by three experienced experts, using 3D reconstructions of femoral grafts. Manual measurements demonstrated excellent intra- and interrater reliability, with ICC values mostly above 0.98 and all exceeding 0.89, validating their use as a reference standard. The automated system showed near-perfect agreement with the expert average (Pearson’s r = 0.99997), with no significant systematic differences observed. Moreover, the automated process required less than 15 seconds per femur with a single user interaction, in contrast to the several minutes and multiple steps needed for manual annotation. The automated measurement system provides accurate, reproducible, and highly efficient assessment of femoral allografts. It eliminates user variability and reduces measurement time by orders of magnitude. The method offers a reliable and scalable solution for integration into routine clinical workflows, and has potential for extension to other anatomical graft types

Project PID2022–137647OB–I00 funded by MCIN/AEI/10.13039/501100011033 and by ERDF, EU

Open Access funding provided thanks to the CRUE-CSIC agreement with Springer Nature