A predictive method for estimating the glenohumeral joint center from palpable landmarks using multiple linear...

Area of Science:

• Biomechanics
• Medical Imaging
• Orthopedics

Background:

• Accurate human motion analysis requires precise tracking of body segments using skin markers.
• Estimating the glenohumeral rotation center (GH-r) is crucial for humerus motion analysis but challenging due to its non-palpable nature.
• Existing linear regression models for GH-r estimation have limitations affecting their performance.

Purpose of the Study:

• To develop an improved linear regression model for estimating the glenohumeral rotation center (GH-r) from palpable landmarks.
• To address and overcome the limitations of current GH-r predictive models.
• To enhance the accuracy and reliability of humerus motion analysis.

Main Methods:

• Utilized a dataset of 73 CT scans, divided into training, validation, and test sets (60:20:20 ratio).
• Developed multiple linear regression models using 4 digitized scapular skin landmarks and subject characteristics as predictors.
• Estimated ground-truth GH-r via spherical fitting of the humeral head; selected the final model based on accuracy and simplicity.

Main Results:

• The final regression model achieved a mean Euclidean distance error (EDE) of 6.81 mm on the test set.
• This represents a statistically significant reduction of at least 10.73 mm compared to established predictive models (p < 0.001).
• Sensitivity analyses indicated a manageable increase in EDE (up to 8.46 mm) with marker placement variability and robustness against inter-observer variability.

Conclusions:

• The developed linear regression model significantly improves the estimation of the glenohumeral rotation center (GH-r).
• This enhanced GH-r estimation offers a more accurate and reliable method for humerus motion analysis.
• The model demonstrates advantages in accuracy and robustness, outperforming existing literature models.