Automatic computation of bernard quadrant in the distal femur

Quantification of the insertion site of anterior cruciate ligament (ACL) in the knee can help understand the role of the ligament and develop its surgical methods to restore its function. The site of ACL insertion could be significantly affected by physician's experience and imaging plane in radiograph. In this study, we proposed a computational method that can detect features on a three-dimensional (3D) model of the distal femur for determining the Bernard quadrant. Forty distal femoral bone models were prepared and their ACL insertion sites were determined by an experienced clinician. Our automatic Bernard quadrant (ABQ) method starts with an algorithm to align a 3D model with a template model using the iterative closest point with scaling method, and to establish the sagittal plane of the distal femur. Then, the Blumensaat's line was calculated based on the intercondylar curve from the sagittal plane view of the femur so that the Bernard quadrant was automatically determined as close as the conventional method. Reliability and accuracy of the proposed method were quantified by comparing the ABQ results against the manual Bernard quadrant (MBQ) performed by three clinicians. The MBQ was determined manually on forty 3D distal femur models and digitally reconstructed radiographs (DRR) from CT images using custom software. ACL insertion site of the forty specimens was located in average at 23.4% along the Blumensaat's line and 24.5% along the height (vertical to the Blumensaat's line) in the ABQ method. Intra-class correlation coefficients (ICCs) for the positions of ACL insertion sites determined in ABQ and MBQ were higher than 0.8 in both parallel and vertical directions to the Blumensaat's line. Intraclass correlation coefficients between the three observers for manually determining ACL insertion site with 3D models was higher than 0.8 and that with DRR models was 0.6. The proposed ABQ method could closely reproduce the ACL insertion sites determined by clinicians in 3D distal femur models. The method would help objective quantification of the ACL insertion site in future clinical ACL research.