Aprendizaje automático asistido para la planificación preoperatoria de artroplastia total de cadera: validación de...

Machine learning (ML) applications in pre-operative templating for total hip arthroplasty (THA) require systematic validation through prediction-verification methodologies to establish clinical utility and safety.

We conducted a retrospective case series of 9 consecutive THA cases between March and September 2025. A hybrid ensemble ML system performed pre-operative templating using standardized radiographs with magnification correction. Predictions were systematically validated against actual intraoperative component selections through post-operative analysis. Primary outcomes included component-specific accuracy with descriptive statistics. Secondary outcomes included clinical safety and feasibility assessment.

The ML system achieved 85.7% overall templating accuracy across all cases. Component-specific accuracies were stem type 100% (9/9), stem sizing 88.9% (8/9), cup sizing 66.7% (6/9), head sizing 54.4% (5/9), and material selection 95.6% (9/9). High templating accuracy (≥97.7%) was achieved in 44.4% of cases (4/9). The system demonstrated conversion risk prediction capability in one case. Clinical outcomes showed a 100% success rate with zero complications and no reoperations at a mean of 17.2 ± 8.1 weeks of follow-up.

ML-assisted pre-operative templating demonstrates feasibility for THA planning with encouraging accuracy results and excellent clinical safety. These findings support the potential for larger validation studies. The prediction-verification methodology provides a systematic framework for ML validation in orthopedic surgery.