Radiographic images of human femora are common tools for diagnosis in clinics. For example, they are used for evaluating the severity of osteoporosis and osteoarthritis. Our aim is to help diagnosis by building new computational methods that can gather more information from these images.
Measurement of 2D bone mineral density (BMD) with dual-energy X-ray absorptiometry (DXA) imaging of proximal femur is the gold standard method for diagnosis of osteoporosis and for predicting fracture risk. However, the DXA images alone are only a moderate predictor of fractures. We believe that with mechanical modeling we can predict the fractures more accurately since then also the geometry of bone is included in prediction of fracture. Therefore, we developed a method to estimate the 3D shape and internal density of the proximal femur based on one DXA image. Then, this estimated shape can be feed for finite element analysis (FEA) to get a subject specific mechanical model and fracture model. From this model we can calculate the maximum strength that the patient's proximal femur can bear without fracture.
The videos present the estimation of the 3D shape and internal density of a proximal femur based on a 2D DXA image. (a) A template which includes general information about femur shape and density distribution is built from femoral CT images. Then, the shape of this template is fitted to the shape of 2D DXA image and the internal density is matched with the one from the DXA image.
A common way to evaluate the geometry of the proximal femur is to take measures from the femoral radiograph, i.e., shape parameters. Then they can be correlated with clinical parameters such as severity of osteoarthritis or osteoporosis to find their relation or to use them as predictor of the clinical parameters. However, each of these shape parameters measure only one local feature of the shape. Therefore, we have built a statistical shape model that can describe the variation in both shape and density of femur in the evaluated population with a few parameters, called modes. The base of the statistical shape model is the principal component analysis. In one application we used the statistical shape model when we created new femur samples from a population. It is possible since the modes describe the variation of the shape and density within the training population (see video). We have also used the statistical shape model for predicting the 3D orientation of the femur in 2D radiographs.
The mode values in the statistical shape model describe the variation of both the shape and internal density in the population’s proximal femora. The value of a mode defines how far the shape is from the average of the population. The unit is standard deviations.
Väänänen Sami P., Isaksson Hanna, Waarsing Erwin, Zadpoor Amir Abbas, Jurvelin Jukka S., Weinans Harrie, Estimation of 3D rotation of femur in 2D hip radiographs, Journal of Biomechanics (2012) accepted for publication.
Functional imaging of knee joint
MRI is commonly used for evaluating condition of knee joint. Since MRI offers relatively good contrast between solid and soft tissues, and furthermore, it offers information about compositions and structures in different joint tissues (cartilage, meniscus), it is used for diagnosing degenerative changes in knee joint, such as cartilage and meniscal injuries [1, 2]. However, MRI does not reveal function of knee joint during different physical activity (walking, running). Therefore, different computational models have been developed estimating varying forces and strains in the knee joint.
In the recent studies [3-5], realistic loading input has been implemented into knee joint models. However, these models have lacks in realistic material characterization for different joint tissues. Cartilages of knee joint have been usually considered as linear isotropic material [3-5], while in reality they have highly anisotropic and time-dependent properties due to depth-dependent collagen fibril network, proteoglycan density and fluid fraction [6-8]. Since all knee joint loading forces are transferred through cartilage tissues to bone, effects of previous compositions should be considered as realistically as possible. This is possible by using fibril-reinforced biphasic properties for cartilage tissues (See Biomechanics and modeling).
For simulating realistically effects of different joint loadings in addition to realistic material characterization, varying knee loadings, rotations and translations (movements of femur respect with to tibia) is needed. This data can be obtained by using different gait analysis methods, such as fluoroscopic setup system or markers [9, 10]. When computational model with realistic material characterizations for different joint tissues and realistic data during physical activity is combined, realistic simulations can be done (11-13). This approach can be used for designing and optimizing treatment operations, such as partial meniscectomy or cartilage/menisucs repair. Furthermore, presented tool can be used for estimating development\probability of osteoarthritis in knee joint.
Comparison of stress variations within knee joint during axial impact loading with different collagen fibril architecture in the tibial cartilage: figures show results of different collagen architectures at the center of medial compartment of the knee joint. In the left figure, collagen orientations are based on the T2-mapped MRI data, whereas in the right figure, collagen orientations are obtained from the literature.
Stress variations in the intact knee joint (top video), in the knee joint with middle radial tear of lateral meniscus (middle video) and in the knee joint with partial lateral meniscectomy during normal walking (bottom video).