Accurate Humeral Version Angle Prediction From Limited Anatomical Data Using Statistical Shape Models.

Area of Science:

• Orthopedic surgery
• Biomedical engineering
• Medical imaging

Background:

• Accurately estimating pre-morbid humeral version angle (HVA) is crucial for successful shoulder replacement surgery.
• Statistical Shape Models (SSMs) are used to reconstruct bone shapes and predict HVA, but require visible humeral anatomy.
• Multisegmented SSMs (MSSMs) leverage inter-bone correlations but their utility for HVA prediction is underexplored.

Purpose of the Study:

• To evaluate the accuracy of a scale-free humeral SSM and a three-shoulder-bone MSSM in estimating HVA.
• To investigate the relationship between HVA accuracy and global/local bone reconstruction quality.
• To compare HVA prediction accuracy between SSM and MSSM under different anatomical visibility scenarios.

Main Methods:

• Sixty-one humeri were segmented from medical images.
• A leave-one-out cross-validation tested SSMs with limited humeral visibility (proximal/distal) and MSSMs with no humeral visibility.
• HVA accuracy was assessed based on reconstruction errors at the marker and global shape levels.

Main Results:

• SSM-based HVA estimation yielded higher accuracy (max error: 3.3° ± 2.3°) compared to MSSM (max error: 6.6° ± 5.4°).
• In both models, prediction errors were primarily influenced by marker-level discrepancies, not global shape fitting.
• The study demonstrated that MSSMs can estimate HVA even without direct humeral data.

Conclusions:

• When predicting anatomical measurements like HVA, prioritizing anatomical accuracy is more critical than achieving the best global shape match.
• SSMs offer superior HVA accuracy when parts of the humerus are visible.
• MSSMs provide a viable alternative for HVA estimation when humeral data is unavailable, leveraging adjacent bone information.