AO classification system and 3D reconstruction view method for radial distal fractures with isolated joint surface collapse bone fragments: An independent inter-observer and intra-observer consistency study

Dong Aozhengzheng, Xing Gao, Yongsheng Li, Zhe Zhao, Wencui Li, Liu Jianquan, Jianwen Yin, Li Jiabei, Cheng Xiangyu, Xiaoqiang Chen, Guanghui WangResearch Square 2024 · DOI

TL;DR

While 3D reconstruction software is often touted as the "gold standard" for surgical planning, this study reveals a nuanced reality: it significantly improves the consistency of locating tiny, isolated bone fragments in wrist fractures, but standard 2D CT scans remain more reliable for high-level fracture classification among experienced surgeons. The findings suggest that 3D modeling is a powerful tool for "mapping" a surgery, but its utility for "categorizing" a fracture depends heavily on a clinician's experience level.

What Problem Does This Paper Address?

Distal radius fractures (common wrist breaks) often involve the "die-punch" mechanism, where the joint surface collapses. A particularly tricky subset involves "isolated fragments"—tiny pieces of bone that have completely detached from the surrounding cortex. If these aren't perfectly repositioned during surgery, the patient faces a high risk of chronic pain and arthritis.

The challenge lies in consistency. Surgeons use the AO (Arbeitsgemeinschaft für Osteosynthesefragen) classification system to communicate and plan treatments. However, interpreting 2D CT slices to determine the exact "zone" of a collapse is difficult. This study investigates whether 3D reconstruction (using MIMICS software) actually improves the reliability of these assessments across different levels of surgical experience.

Methods

The researchers conducted a retrospective consistency study using imaging from 80 patients treated at Shenzhen Second People’s Hospital.

  • The Assessors: Two senior orthopedic clinicians and two junior clinicians.
  • The Workflow: Each assessor evaluated the 80 cases twice, six weeks apart, to measure both inter-observer (agreement between different people) and intra-observer (agreement with oneself over time) consistency.
  • The Tech: They compared standard 2D CT scans (coronal, sagittal, and axial) against 3D models generated by MIMICS 26.0 software.
  • The Metrics: Reliability was measured using the Intraclass Correlation Coefficient (ICC). They specifically looked at AO classification and a proposed "six-zone method" for locating fragments (e.g., scaphoid fossa palmar vs. lunate fossa dorsal).

Datasets

The study utilized a primary dataset of 80 clinical cases from Shenzhen Second People's Hospital, spanning November 2022 to February 2024. All patients had Type B or C distal radius fractures involving articular collapse. No external or public benchmarks were used.

Key Insights and Ideas

  • The "2D Native" Advantage: Surprisingly, senior doctors were more consistent when using traditional 2D CT scans for AO classification. This suggests that veteran surgeons have "trained" their brains to mentally reconstruct 3D structures from 2D slices, and the introduction of a software-generated 3D model actually disrupted their established diagnostic workflow.
  • 3D as a "Junior" Equalizer: Junior doctors showed higher consistency in classification when using 3D reconstructions. For less experienced clinicians, the intuitive nature of a 3D model bypasses the steep learning curve of interpreting 2D spatial data.
  • Mapping vs. Categorizing: While 3D models didn't necessarily help with high-level AO classification, they were vastly superior for localization. Agreement on exactly where a fragment was located reached "almost perfect" levels with 3D reconstruction.
  • Hidden Connections: The study found that fragments appearing "isolated" on 2D slices were sometimes revealed to be partially connected to the bone cortex in 3D. This is a critical distinction: connected fragments can often be repositioned indirectly, while truly isolated ones require more invasive direct exposure.

Key Results

  • AO Classification (Inter-observer): 2D CT scans actually yielded higher consistency (0.642) than 3D reconstructions (0.577).
  • Fragment Localization (Inter-observer): 3D reconstruction was the clear winner. In the coronal plane, 3D consistency was 0.806 (vs. 0.762 for 2D); in the sagittal plane, it was 0.809 (vs. 0.698 for 2D).
  • Reliability Levels: According to the Landis and Koch scale, 3D reconstruction moved fragment positioning from "substantial agreement" to "almost perfect agreement."

Limitations and Future Work

The authors acknowledge that their sample size of 80 is relatively small. Furthermore, the study focused exclusively on the AO classification system; other systems, like the Melone classification (which focuses more on the articular surface), were not tested.

Crucially, this study measured consistency, not accuracy. We know the surgeons agreed more when using 3D for localization, but the study doesn't definitively prove that the 3D location was the "truth" (which would require intraoperative confirmation). Future work should link these imaging methods directly to surgical outcomes and recovery times to see if the increased consistency translates to better patient health.

Summary updated

Abstract

<title>Abstract</title> Purpose This study aimed to compare the consistency among surgeons of varying experience levels in classifying distal radius articular collapse fractures using conventional CT scans and 3D images reconstructed with software MIMICS 26.0 (MIMICS, Boston, MA)according to the AO classification. Additionally, the study compared the consistency in identifying the position of isolated bone fragments on the articular surface using both imaging methods. The aim was to determine a more practical approach for the AO classification of distal radius articular collapse fractures and identifying isolated bone fragment positions, thereby assisting clinicians in preoperative decision-making. Methods The study included 80 patients with distal radius articular collapse fractures. Four independent assessors (two senior and two junior orthopedic clinicians) evaluated the fractures' AO classification and the relative positions of isolated articular bone fragments using conventional CT images and 3D reconstructed images. After six weeks, the 80 cases were randomly reassigned, and all assessors re-evaluated the abovementioned parameters. The intraclass correlation coefficient (ICC) was used to calculate inter-observer and intra-observer reliability, presented with a 95% confidence interval. Results The results showed that AO classification of distal radius articular collapse fractures using 3D reconstructed images had lower overall inter-observer consistency (0.577 (0.369–0.729)) compared to direct AO classification from conventional CT scans (0.642 (0.486–0.761)). However, determining the relative positions of bone fragments in the coronal plane using 3D reconstructed images showed higher overall inter-observer consistency (coronal plane 0.806 (0.700–0.882); sagittal plane 0.809 (0.712–0.879)) compared to conventional CT scans (coronal plane 0.762 (0.635–0.852); sagittal plane 0.698 (0.558–0.802)). Conclusion In conclusion, for distal radius articular collapse fractures, the consistency in AO classification assessments among observers was generally moderate, influenced by the experience level of the doctors. Junior doctors showed higher consistency in classification assessments using 3D reconstruction, while senior doctors had higher consistency using conventional CT scans. The method of observation for bone fragment positions, whether by conventional CT scan or 3D reconstruction of the distal radius, showed differences in consistency, with 3D reconstruction reaching almost perfect consistency levels in identifying bone fragment positions on the articular surface.
BibTeX
@inproceedings{aozhengzheng2024ao,
  author = {Dong Aozhengzheng and Xing Gao and Yongsheng Li and Zhe Zhao and Wencui Li and Liu Jianquan and Jianwen Yin and Li Jiabei and Cheng Xiangyu and Xiaoqiang Chen and Guanghui Wang},
  title = {{AO classification system and 3D reconstruction view method for radial distal fractures with isolated joint surface collapse bone fragments: An independent inter-observer and intra-observer consistency study}},
  booktitle = {Research Square},
  year = {2024},
  doi = {10.21203/rs.3.rs-4588802/v1},
}
Figure