Beyond Benchmarks: Towards Robust Artificial Intelligence Bone Segmentation in Socio-Technical Systems

Abstract

We present a comprehensive multi-center evaluation of 20 state-of-the-art bone segmentation models across 1,000 clinically resampled CT/CBCT scans. Our study reveals that image sharpness, isotropic smaller voxels, and neutral orientation significantly improve segmentation performance, while metallic osteosynthesis and anatomical complexity lead to significant degradation. The findings highlight the gap between benchmark performance and real-world clinical applicability, emphasizing the need for robustness evaluation in socio-technical systems.

Kunpeng Xie

Oral & Maxillofacial Surgeon & Researcher

Doctoral candidate at RWTH Aachen University, developing clinically grounded tools for image analysis, computer-aided surgery, and augmented reality navigation.