Patient‐specific parameterised cam geometry in finite element models of femoroacetabular impingement of the hip

Background: Impingement resulting in soft tissue damage has been observed in hips with abnormal morphologies. Geometric parameterisation can be used to automatically generate a range of bone geometries for use in computational models, including femurs with cam deformity on the femoral neck. Methods: This study verified patient‐specific parametric finite element models of 20 patients with cam deformity (10 female, 10 male) through comparison to their patient‐specific segmentation‐based equivalents. The parameterisation system was then used to generate further models with parametrically defined geometry to investigate morphological changes in both the femur and acetabulum and their effects on impingement. Findings: Similar findings were observed between segmentation‐based and parametric models when assessing soft tissue strains under impingement conditions, resulting from high flexion and internal rotations. Parametric models with cam morphology demonstrated that clinically used alpha angles should not be relied on for estimating impingement severity since planar views do not capture the full three‐dimensional geometry of the joint. Furthermore, the parametric approach allowed study of labral shape changes, indicating higher strains can result from bony overcoverage. Interpretation: The position of cams, as well as their size, can affect the level of soft tissue strain occurring in the hip. This highlights the importance of reporting the full details of three‐dimensional geometry used when developing computational models of the hip joint and suggests that it could be beneficial to stratify the patient population when considering treatment options, since certain morphologies may be at greater risk of elevated soft tissue strain.


INTRODUCTION
Abnormal bone morphology in the hip is associated with femoroacetabular impingement 47 (FAI), in which repeated contact between the proximal femur and the acetabular rim can 48 result in pain and intra-articular damage [1]. A particular example is cam deformity, in which 49 excess bone is present on the femoral neck. Cams most typically occur in young adults, and 50 are more prevalent among males [2]. Understanding of the circumstances leading to 51 symptomatic impingement remains elusive, especially because some hips possessing 52 morphology characteristic of FAI remain asymptomatic [3]. 53 In order to investigate the effects of bone morphology on tissue strains computationally, it 54 is useful to be able to automatically generate multiple geometries representative of the 55 population variation. This can be achieved using a parametric approach to finite element 56 models of the hip [4,5]. A recent study [4] demonstrated that parameterised models could 57 identify differences in contact mechanics between two different subjects with healthy hips 58 across a gait cycle, providing confidence that such models can be used to systematically 59 3 evaluate the effects of clinically relevant changes in morphology. However, some studies 60 suggest that models with idealised geometry can lead to poor estimates of hip contact 61 stresses [6,7]. It is therefore important that parametric models are compared with 62 segmented patient-specific models in order to understand the effects of smoothing out local 63 undulations in subject-specific articular geometries. As well as isolating the effects of 64 individual changes, parametric models with simplified articular surfaces can alleviate 65 computational convergence issues [8] reported to occur when using more complex 66 geometry [9]. 67 Geometrical variations generated in parametric models must be well defined. Clinically used 68 radiographic measurements such as the alpha angle, which estimates the asphericity of the 69 femoral head, are highly dependent on the two-dimensional radiographic view of the joint 70 and do not capture the full three-dimensional geometry [10,11]. Alpha angles can therefore 71 be ambiguous and are not well suited to describing geometrical variation.  The aims of this study were to:

86
We previously developed a geometric parameterisation system capable of representing 87 segmented femurs with cam deformity with root mean squared surface fitting errors in the 88 region of 0.6 mm, allowing isolation of the size and position of cams [10]. The 89 parameterisation method allowed generation of new femoral geometries with the neck 90 region described by ellipses (Fig. 1    Generation of all models was automated in Abaqus 6.14 (Dassault Systèmes, Vélizy-124 Villacoublay, France) using Python. All FE models were quasi-static analyses, with geometric 125 non-linearity.   In all models in this study, femoral bones and the acetabulum were modelled as rigid bodies The mesh density adopted was determined after mesh convergence tests. Displacements

197
The data associated with this paper are openly available from the University of Leeds data particularly > 1 mm. (Fig. 5).

211
High deformation of elements in the cartilage-labrum junction prevented models converging 212 past a certain level of rotation, so for the additional parametric tests, rotation levels where 213 all models converged were used to generate comparison graphs. This was 15 o for the 214 7 models where cam morphology was varied (Fig. 6) and 25 o for the models where acetabular 215 rim morphology was varied (Fig. 7).

216
When cam morphology was varied, measured alpha angles did not predict outputs of 217 interest (Fig. 6). In particular, AP alpha angles were unexpectedly higher ( 63. When acetabular rim and labral morphology was varied, an increase to bone coverage had 226 the greatest effect on impingement severity (Fig. 7). A 10% increase in bone (with labrum

299
In the acetabular coverage tests, greater bony coverage resulted in increased strain (Fig. 7).

300
The models therefore predicted that elevated bony acetabular coverage likely increases 301 impingement severity for a given level of rotation. Labral displacement appeared to be 302 driven by the position of its tip relative to the cam, rather than overall labral length. The