Automated image registration for assessing three-dimensional alignment of entire lower extremity and implant position using bi-plane radiography.

Abstract

An automated image-matching technique is presented to assess alignment of the entire lower extremity for normal and implanted knees and the positioning of implants with respect to bone. Sawbone femur and tibia and femoral and tibial components of a total knee arthroplasty system were used. Three spherical markers were attached to each sawbone and each component to define the local coordinate system. Outlines of the three-dimensional (3D) bone models and component computer-aided design (CAD) models were projected onto extracted contours of the femur, tibia, and implants in frontal and oblique X-ray images. Three-dimensional position of each model was recovered by minimizing the difference between the projected outline and the contour. Median values of the absolute error in estimating relative positions were within 0.5mm and 0.6 degrees for the femur with respect to the tibia, 0.5mm and 0.5 degrees for the femoral component with respect to the tibial component, 0.6mm and 0.6 degrees for the femoral component with respect to the femur, and 0.5mm and 0.4 degrees for the tibial component with respect to the tibia, indicating significant improvements when compared to manually obtained results.