Morphological errors in 3D bone models of the distal femur and proximal tibia generated from magnetic resonance imaging and computed tomography determined using two registration methods

Abstract

ABSTRACTThree dimensional models of the femur and tibia from MRI and CT are used in various studies of knee biomechanics but morphological errors occur. The objectives were to determine the morphological errors in 3D bone models from MRI and CT using clinically-adopted imaging protocols, to determine whether these errors are affected by the method of registration to a gold-standard 3D model, and to determine whether the slice thickness of CT affects errors. CT and MRI scans of 13 human cadaveric knees were performed and 3D bone models of the distal femur and proximal tibia were compared to a gold-standard 3D bone model generated by laser scanning. Registration of the models from MRI and CT to those from laser scanning was performed using the iterative closest point (ICP) algorithm and fiducial markers. The key findings were that 1) 3D bone models generated with CT yield submillimeter accuracy while 3D models generated with MRI yield accuracy worse than 1 mm; 2) CT bone models slightly overestimate the bone morphology, while the MRI bone models substantially underestimate it; 3) morphological errors of 3D bone models from 0.625 mm and 1.25 mm slice thickness CT scans are not different; and 4) ICP registration underestimates the morphological errors compared to registration with fiducial markers.