A method for registration of full-limb radiographs to knee MRI

Abstract

Knee osteoarthritis (OA) is a major source of morbidity in the aging population with a prevalence of over 16 % in people older than 45 [1]. Previous studies have correlated lower extremity alignment with disease progression [2]. This is defined by the mechanical axis, which is the vector running from the center of the hip joint to the center of the talar dome and can be measured on weight-bearing full limb radiographs [2]. Varus alignment, when the mechanical axis passes medial to the central point between the tibial spines, has been linked to progression of structural damage in the medial compartment of the knee, while valgus alignment, when the mechanical axis passes lateral to the central point between the tibial spines, has been linked to progression of lateral compartment disease [3, 4]. Thus, evaluation of a patient’s lower extremity alignment is important in the assessment of knee OA and treatment planning. Another principal imaging modality in knee OA, magnetic resonance (MR) imaging, has become increasingly utilized for the assessment of joint damage [5, 6]. To improve our understanding of the pathogenesis of knee OA, it would be useful to directly compare the location of damage in relation to the mechanical axis in the knee joint. However, it is not possible to localize the mechanical axis on a clinical knee MRI as the hip and ankle joints are not included in the imaging field of view (FOV). In addition, most MR imaging is not performed with the patient bearing weight, and full-limb magnetic resonance imaging is not often performed. Approximations of these alignment angles can be measured using limited FOV radiographs as well as the portion of the knee visualized on knee MRI. However, studies using radiographs limited to the knees showed that these measurements have limited correlation with the true mechanical axis [7]. To determine the location of the mechanical axis relative to damage identified on knee MRI, we present a method of registering full-limb radiographs to coronal knee MR images. We also present as an example, implementation of this method for quantification of the spatial relationship between the mechanical axis and areas of bone marrow edema-like signal on knee MR images.