A Novel Method for the Approximation of Humeral Head Retrotorsion Based on Three-Dimensional Registration of the Bicipital Groove

Background: The accurate restoration of premorbid anatomy is key for the success of reconstructive surgeries of the proximal part of the humerus. The bicipital groove has been proposed as a landmark for the prediction of humeral head retrotorsion. We hypothesized that a novel method based on bilateral registration of the bicipital groove yields an accurate approximation of the premorbid anatomy of the proximal part of the humerus. Methods: Three-dimensional (3D) triangular surface models were created from computed tomographic data of 100 paired humeri (50 cadavers). Segments of the distal part of the humerus and the humeral shaft of prespecified lengths were defined. A surface registration algorithm was applied to superimpose the models onto the mirrored contralateral humeral model based on the defined segments. We evaluated the 3D proximal humeral contralateral registration (p-HCR) errors, defined as the difference in 3D rotation of the humeral head between the models when superimposed. For comparison, we quantified the landmark-based retrotorsion (LBR) error, defined as the intra-individual difference in retrotorsion, measured with a landmark-based 3D method. Results: The mean 3D p-HCR error using the most proximal humeral shaft (bicipital groove) segment for the registration was 2.8 degrees (standard deviation [SD], 1.5 degrees; range, 0.6 degrees to 7.4 degrees). The mean LBR error of the reference method was 6.4 degrees (SD, 5.9 degrees; range, 0.5 degrees to 24.0 degrees). Conclusions: Bilateral 3D registration of the bicipital groove is a reliable method for approximating the premorbid anatomy of the proximal part of the humerus.