Reconstruction of Scapula Bone Shapes from Digitized Skin Landmarks Using Statistical Shape Modeling and Multiple Linear Regression.

PURPOSE

The aim of this study was to develop an algorithm for the reconstruction of scapula bone shapes from skin landmarks, using a statistical shape model (SSM).

METHODS

A sample of 56 scapula segmentations was used, as well as 4 scapular bone and skin landmarks. Regression models were built to predict the coordinates of bone landmarks from skin landmarks using subject-specific variables, namely skin landmark coordinates, sex, age, weight, and height. The scapula shapes were reconstructed by fitting the bone landmarks of the SSM's mean shape to the predicted bone landmarks of the subject.

RESULTS

The developed regression models registered a R ranging from 0.70 to 0.98, with a maximum median error of 4 mm. The average surface-to-surface errors were equal to 2.41 and 2.45 mm using digitized and predicted bone landmarks, respectively. No significant statistical differences were observed between scapula shapes reconstructed from digitized and predicted bone landmarks.

CONCLUSION

This study demonstrated the reliability of the developed algorithm in deriving subject-specific scapula shapes from experimentally acquired data, highlighting that scapula shape reconstructions based on a limited set of landmarks can effectively generate subject-specific computational models without the need for additional medical imaging.