Knee 3D reconstruction from radiographs: systematic review finds 0.5-1.0 mm accuracy in recent methods

Systematic review of 28 studies: 3D knee reconstruction from 2D radiographs now achieves 0.5–1.0 mm root-mean-square error. Accuracy often adequate with two orthogonal views. Smaller validation cohorts and limited pathology representation remain key limitations.

• Twenty-eight eligible studies evaluated, spanning statistical shape modeling, deep learning, model-based registration, and hybrid approaches.
• Recent methods achieve approximately 0.5–1.0 mm root-mean-square error, with processing times from under one second to several minutes.
• Primary limitations include small validation cohorts, limited pathological representation, and reliance on simulated or digitally reconstructed radiograph validation.

Automated summary

RadPigeon summaries are original and for information only. They are not clinical advice.