Prediction of the subject-specific knee passive motion from non-invasive measurements

The passive motion of the knee is the natural motion of the joint in unloaded conditions. It is the joint starting condition when loads are applied, thus affecting the joint behaviour also in loaded conditions. For this reason, the knowledge of this motion is useful in all applications which aim at replicating or restoring the natural behaviour of the knee, such as lower-limb modelling, surgical planning, prosthetic design. Several studies measured the passive motion of the joint and the mean results can be used as a reference [1-3]. However, there is an increasing request of subject-specific results that would allow personalization of model parameters or of prosthesis geometry on a patient. In these cases, the subject motion would be required. An accurate estimation of the joint motion is difficult to obtain in vivo: non-invasive techniques can be inaccurate (skin-markers) or too complicated (fluoroscopy) for standard practice, while more invasive techniques (bone-pins) are not acceptable in several applications. Thus, new techniques are needed to predict the natural joint motion with a good accuracy, starting from non-invasive measurements. Two techniques were proposed for the knee passive motion modelling. Both methods start from the consideration, supported by relevant experimental analyses [2,3], that in passive conditions the knee behaves as a single degree-of-freedom (1DOF) system guided by the passive structures of the joint. The first technique (T1) predicts the passive motion by maximizing the joint congruence at all flexion angles [4]: it only requires the 3D model of articular surfaces with menisci, that can be obtained by MRI. T1 was developed for the ankle joint. At this stage, its application to the knee revealed a good accuracy, but the errors increased at high flexion angles, particularly for some motion components. The second technique (T2) models the knee as a 1DOF spatial mechanism, featuring the two articular contacts and the three isometric fibres of the anterior cruciate, posterior cruciate and medial collateral ligaments [3,5]. T2 was very accurate to replicate the passive motion of specimens over the full flexion arc, but a reference motion is needed to adjust the model parameters and the specimen motion was previously used [3]. The limitations of T1 and T2 are overcome in this paper by combining the two methods. An estimate of the joint motion is obtained by T1. This estimate is used as an input for T2 by which a new motion prediction is defined. The idea is that ligament constraints of T2, starting from the estimate provided by T1, can improve the overall motion prediction over the full flexion arc. The technique is here presented and applied on a specimen.