Work

A Comprehensive Computational Workflow for Patient-Specific 3D Reconstruction and Quantitative Analysis of Acetabular Revision Defects

Year: 2025

Type: article

Abstract: Quantitative analysis of acetabular bone loss in hip revision surgery remain limited. This hinders evidence-based implant selection and the development of standardised defect models. We present a computational workflow to reconstruct patient-specific revision cases as three-dimensional (3D) defect volume models to derive intersections and multi-sphere models. A total of 47 acetabular revision cases (Paprosky IIA–IIIB) were processed retrospectively. The acetabular defect surface mesh was isolated using curvature-guided manual refinement (mean boundary error: 2.62 ± 1.69 mm), followed by distortion-free scaling and weighted alignment to a statistical shape model (SSM) of the pelvis. Two strategies were implemented to define the defect volumes: a fast Alpha Method and a high-fidelity Hybrid Model, which supplements the defect model with interior points of the reference pelvis. Threshold-based intersections, spatial defect distribution, and multi-sphere approximations (up to 10 spheres) were generated and benchmarked. The Hybrid Models preserved 77.26 % ± 21.55 % of the original acetabular defect vertices (alpha radius 6.23 ± 4.55), compared to 34.04 % ± 20.26 % for the Alpha Method (alpha radius 32.53 ± 9.97). The intersection analyses revealed a relevant, continuous acetabular margin, which Alpha Method failed to detect. The representative multi-sphere defect models achieved a mean surface error of less than 5 mm. This workflow creates representative 3D volume models of acetabular defects and compact multi-sphere models using standard computed tomography (CT) data. It provides a reproducible, quantitative framework for the objective classification and analysis of defects. It can also support implant planning and provide a standardised defect model for use in experimental or finite element studies.

Source: IEEE Access

DOI: 10.1109/access.2025.3626705