Incorporating Patient-Specific Hip Orientation from Weightbearing Computed Tomography Affects Discrete Element Analysis-Computed Regional Joint Contact Mechanics in Individuals Treated with Periacetabular Osteotomy for Hip Dysplasia

Abstract

Computational models of the hip often omit patient-specific functional orientation when placing imaging-derived bony geometry into anatomic landmark-based coordinate systems for application of joint loading schemes. The purpose of this study was to determine if this omission meaningfully alters computed contact mechanics. Discrete element analysis models were created from non-weightbearing (NWB) clinical CT scans of 10 hip dysplasia patients (11 hips) and oriented in the International Society of Biomechanics (ISB) coordinate system (NWB-ISB). Three additional models were generated for each hip by adding patient-specific stance information obtained via weightbearing CT (WBCT) to each ISB-oriented model: (1) patient-specific sagittal tilt added (WBCT-sagittal), (2) coronal and axial rotation from optical motion capture added to (1) (WBCT-combo), and (3) WBCT-derived axial, sagittal and coronal rotation added to (1) (WBCT-original). Identical gait cycle loading was applied to all models for a given hip, and computed contact stress and contact area were compared between model initialization techniques. Addition of sagittal tilt did not significantly change whole-joint peak (p=0.922) or mean (p=0.871) contact stress or contact area (p=0.638). Inclusion of motion-captured coronal and axial rotation (WBCT-combo) decreased peak contact stress (p=0.014) and slightly increased average contact area (p=0.071) from WBCT-sagittal models. Including all WBCT-derived rotations (WBCT-original) further reduced computed peak contact stress (p=0.001) and significantly increased contact area (p=0.001). Variably significant differences (p=0.001 to 1.0) in patient-specific acetabular subregion mechanics indicate the importance of functional orientation incorporation for modeling applications in which local contact mechanics are of interest.

Graphical Abstract

Introduction

Hip dysplasia is a musculoskeletal condition that results in insufficient coverage of the femoral head by the acetabulum. This lack of coverage leads to hip pain and accelerated progression of osteoarthritis [1], presumably from instability and abnormalities in the force distribution within the hip joint [2]. Computational modeling techniques such as finite element analysis (FEA) and discrete element analysis (DEA) are commonly used to evaluate variations in cartilage contact stress that result from abnormal force distribution in a variety of joints [3–5]. Importantly, such computational techniques have been shown to provide reliable and accurate contact stress calculations in cadaveric validation studies performed in the hip [6, 7].

Computational models of living patients are typically generated from clinical, non-weightbearing supine CT or MRI scans, and the anatomy derived from these non-weightbearing scans must be reoriented to a neutral, standardized pose before functional load application. This reorientation process often involves rotating and translating the 3D anatomy models to align with a standardized coordinate system defined by bony landmarks, and then applying joint rotations and loading vectors defined within that same coordinate system. While multiple hip joint coordinate systems exist for definition of a known functional orientation [8–13], application of any one of them effectively standardizes model orientation between patients. This removes critical elements such as patient-specific femoral version and pelvic tilt during weightbearing from the resulting computational models [14–17]. Standing pelvic tilt derived from two-dimensional weightbearing radiographs has been incorporated into a limited number of patient-specific FEA models, resulting in computation of decreased hip contact area, and increased median contact pressure [14]. However, even with this technique [14, 18], other potentially relevant features of an individual’s acetabular-femoral relationship (hip extension/abduction) are omitted.

Similar to how weightbearing radiographs visualize the joint in a loaded position, thereby facilitating a more accurate assessment of functional acetabular coverage, femoral version, and joint space narrowing [15, 19], the recently available technology of weightbearing CT (WBCT) similarly captures functional joint position in 3D while a patient is standing in a loaded, functional orientation. To date, most work with standing WBCT has focused on evaluating joint space distance [20], congruency and coverage [21]; characterizing deformities [22]; or defining alignment [23] in the foot, ankle, and knee. Such studies have found significant differences between measurements made on scans acquired in a WBCT versus a conventional supine (non-WBCT) CT scanner [24–27]. For example, Richter, et. al. found an average difference of 9 degrees associated with weightbearing when evaluating the 1^st – 2^nd intermetatarsal angle in the foot [27]. Hirschmann, et. al. found average differences of 1 mm when assessing the medial femorotibial joint space width in the loaded versus unloaded knee [24].

Beyond providing relatively simplistic measurements of joint alignment, this emerging technology holds the exciting potential for simultaneously providing 3D geometry and functional position data for generating more accurate patient-specific computational models of joints. However, while several studies have utilized WBCT scans for computational model development in other joints [20, 22–24, 28, 29], WBCT scans have not been used to create computational models of the hip. Though improved patient-specificity of computed contact mechanics at the hip are appealing, the effects of incorporating WBCT data on the computed contact mechanics are unclear. Therefore, the purpose of this study was to evaluate how joint contact mechanics computed from models based on an upright, WBCT-derived, patient-specific orientation differed from mechanics computed from models developed by placing clinical CT-derived geometry into a standardized, anatomic landmark-based orientation. We hypothesized that inclusion of a patient-specific weightbearing pose would result in calculation of higher contact stress and differences in contact area in different locations within the acetabulum.

Methods

With University of Iowa Institutional Review Board approval (IRB ID: 202002382), 10 patients (11 hips) indicated for periacetabular osteotomy (PAO) to treat hip dysplasia between June 2020 and August 2021 consented in writing to undergo a preoperative, research-specific pelvic WBCT scan (HiRise; Curvebeam, Hatfield, PA) (effective dose: 0.92 mSv; 0.3 – 0.5 mm isotropic voxels; 1300×1300 in-plane resolution) (Figure 1). Indications for PAO were lateral center edge (LCEA) ≤20 degrees or a LCEA from 20–25 degrees with high internal rotation in flexion and persistent hip pain for >6 months despite adequate rehabilitation and activity modification. Patients (9 female) were an average of 18.7 ± 4.2 years of age and weighed an average of 62.2 ± 5.8 kg (BMI 22.9± 2.2) at the time of surgery. Three hips had LCEA in the range of borderline dysplasia (22.7°±1.5°), and the other 8 hips had mild/moderate dysplasia (LCEA 17.6°±2.6°). LCEA was measured on each patient’s standing anteroposterior diagnostic 2D radiographs [30] by the treating board-certified orthopedic surgeon. Standard-of-care clinical pelvis CT scans (effective dose: 1.05 mSv; 0.6 – 0.7 mm isotropic voxels; 512×512 in-plane resolution) were also obtained from each patients’ medical record.

Figure 1:

A clinical, non-weightbearing CT scanner (A) is used to capture bony geometry in high resolution (B) while a patient is lying supine in a non-functional orientation. A weightbearing CT scanner (C) captures the bony geometry (D) while the patient is in a functional standing orientation.

Model generation technique

Femoral and pelvic bony geometries were segmented from non-WBCT and WBCT scans in Mimics (Materialise, Leuven, Belgium). The resulting triangulated surface models were smoothed to reduce any stairstep artifact present in the segmentation (GeoMagic Design X; 3D Systems, Inc., Rock Hill SC) [31]. Due to the inability to visualize articular cartilage on clinical CT scans, cartilage was approximated by manually defining the subchondral bone surfaces of the femoral head and acetabular lunate on the non-WBCT surface models and projecting those surfaces a distance corresponding to half of the patient’s joint space. Any resulting local overlaps were back-projected prior to application of a custom smoothing algorithm, which permitted a maximum radial change of 0.5 mm in the projected cartilage surface. The result was a smooth representation of continuous layers of nonuniform-thickness acetabular and femoral head cartilage [31].

Model Alignment –NWB-ISB Models

Following well-established methods for preparing clinical imaging-derived hip models for simulation of walking gait [14, 18, 31, 32], the pelvis and femur geometry obtained from each patient’s non-WBCT scan were aligned to the standardized International Society of Biomechanics (ISB) hip joint coordinate system using manually identified anatomic landmarks (Figure 2) [12]. Briefly, the medial-lateral pelvic axis (z_p axis) was defined as the line passing through the center of a best-fit sphere to the acetabular lunate of the hip being modeled and parallel to a line connecting the left and right anterior superior iliac spines (ASISs). The superior-inferior axis (y_p axis) was composed of the line passing through the center of the acetabulum and perpendicular to the plane defined by the left and right ASISs and the midpoint of the posterior superior iliac spines. The anteroposterior axis (x_p axis) was defined perpendicular to both the y_p and z_p axes and passing through the center of the acetabulum. The pelvis was reoriented to align the local pelvic (x_py_pz_p) coordinate system to a global (XYZ) coordinate system such that the positive X axis points anteriorly, the positive Y axis points superiorly, the positive Z axis points medially from the left hip (laterally from the right hip), and the origin is centered at the center of the acetabulum.

Figure 2:

For each hip, the non-WBCT bony geometry was used in all 4 DEA calculations to isolate differences in computed contact stress to be only the result of differences in initial model position, rather than differences associated with segmentation on different CT scans. To create each patient’s NWB-ISB model (left), the non-WBCT geometry (blue) was manually reoriented by selecting bony landmarks and rotating/translating the bony surfaces to bring those landmarks into alignment with the International Society of Biomechanics coordinate system. To create the basis for the 3 different WBCT models (right), a copy of the non-WBCT geometry was spatially registered to the WBCT segmentation (yellow) and used as the baseline WBCT geometry (gray).

The femur was placed into the ISB coordinate system by first fitting a cylinder to the femoral condyles and defining the medial and lateral femoral epicondyles as the locations where the cylinder axis exited the bone. The superior-inferior axis (y_f axis) was defined by the line connecting the midpoint of the medial and lateral epicondyles to the center of a best-fit sphere to the femoral head. The anterior-posterior axis (x_f axis) was defined as an anteriorly directed line passing through the center of the femoral head and perpendicular to the plane defined by the medial/lateral epicondyles and the femoral head center. Finally, the medial-lateral axis (z_f axis) was defined through the center of the femoral head and perpendicular to both the x_f and y_f axes. The femur was reoriented to align these local femoral axes with the same global XYZ coordinate system as the pelvis.

Model Alignment – Standing WBCT Models

A copy of the original, non-WBCT geometry was spatially registered to the segmented WBCT surfaces using an iterative closest point (ICP) algorithm in GeoMagic Design X (Figure 2). This reoriented surface geometry was subsequently used in WBCT model construction because using identical geometry allowed for identifying differences in computed contact stress that resulted only from differences in initial position rather than also from differences in segmentation based on imaging modality.

The Z-axis of the WBCT model in the original scanner coordinate system approximated the superior-inferior axis, while the Z-axis was medial/lateral in the ISB coordinate system (Figure 3). Therefore, the WBCT model was provisionally reoriented to make the ASIS midpoint of the WBCT model coincident with the ASIS midpoint on the NWB-ISB model and align the superior-inferior axis of the WBCT model with the ISB superior-inferior axis. The WBCT model was then rotated in the axial plane around the superior-inferior axis (now the vertical Y-axis) through the ASIS midpoint until a bony landmark-defined anterior-posterior axis on the WBCT model pointed in the ISB-defined anterior-posterior direction, thus ensuring that both models were facing in the same direction, but all other elements of patient-specific standing (pelvic tilt, hip flexion, etc.) were preserved. Next, an ICP algorithm was used to calculate the transformation matrices required to align these provisionally oriented WBCT bone surfaces to their respective ISB-aligned pelvis and femur surfaces. A ZYX Euler angle decomposition of those transformation matrices provided patient-specific femoral and pelvic rotations associated with 2-legged stance in the WBCT scanner.

Figure 3:

Derivation of WBCT stance orientation was a stepwise process that provided initialization conditions for the WBCT models. In step 1, the WBCT model (gray) was provisionally aligned by translating it such that the ASIS midpoint matched that of the NWB-ISB model (blue). Step 2 was rotating the WBCT model around the ASIS midpoint until the vertical (superior-inferior) axis matched the superior-inferior axis of the ISB coordinate system. For step 3, the bony landmarks for the ISB coordinate system were identified on the WBCT model, and an anterior-posterior (AP) axis was constructed. The model was rotated about the superior-inferior axis until the AP axis defined on the WBCT model was coincident with the AP axis on the NWB-ISB model in an axial view. Finally, in step 4, an ICP algorithm was used to calculate the transformation matrix needed to align the WBCT bone surface to the NWB-ISB bone surface, and an ZYX Euler angle decomposition was used to calculate rotation angles about the principal ISB axes.

A two-legged stance orientation is not the neutrally oriented ISB reference frame from which to apply gait-related rotations [33]. Therefore, a standing corrective factor was added to the patient-specific femur and pelvis angles to obtain an orientation from which to apply literature-derived walking gait loading data. As previous kinematic studies have demonstrated there are minimal differences between the stance pose of individuals with hip dysplasia compared to individuals with normal hip anatomy [34], angular values comprising average “generic standing” were defined from kinematics obtained from optical motion capture of 9 normal individuals without hip pathology (4 female; 22 ± 1.7 years old; 74.0 ± 13.2 kg) performing a self-selected double-leg stance trial as part of other IRB-approved protocols at our institution. The resulting angular values comprising average generic standing were an average 10°/0.10°/0.08° of pelvic anterior tilt, axial rotation, and coronal rotation, respectively; and 8°/4°/5° of femoral extension, external rotation, and femoral abduction, respectively. These generic standing angles were subtracted from each patient’s WBCT-derived bone orientation angles (omitting the pelvic axial and coronal rotational corrections of less than 1°), thereby leaving a set of patient-specific pelvis and femur offset angles to be added to the joint angles associated with walking gait during DEA modeling. This offset to adjust for a two-legged image acquisition posture left the patient-specific functional neutral stance information as the only difference between the NWB-ISB and WBCT orientation.

However, during this process, it was noted that the WBCT-derived patient-specific femoral rotation in the axial (22.4 ± 8.2° internal rotation) and coronal (3.9 ± 1.6° adduction) planes were abnormal, and likely an artifact of foot positioning on the narrow WBCT platform (Figure 4). Therefore, three different sets of WBCT models were developed to systematically explore the addition of progressively more patient specificity derived from WBCT scanning. Each set of models used identical bony geometry but was initialized into a different position for the application of a loading gait cycle during DEA (Figure 5). First, a set of WBCT-sagittal models was developed in which only the sagittal plane offset rotations of both the femur and pelvis were added to the patient’s NWB-ISB model. Second, to reinclude some effect of hip orientation in the axial and coronal plane, another set of WBCT models (WBCT-combo) was developed that added motion-captured generic stance axial (4° external rotation), and coronal (5° abduction) rotation, plus WBCT patient-specific sagittal rotation to the NWB-ISB models. This allowed for a systemic study of the inclusion of realistic axial and coronal rotation compared solely to sagittal rotation. Lastly, a set of models was developed that utilized all elements of the WBCT-derived patient-specific axial, coronal, and sagittal plane rotation in the initial model positioning (WBCT-original). Through comparison with the WBCT-combo results, this WBCT-original group allowed for the evaluation of differences in contact stress patterns that resulted from hip rotation while standing in the confines of the WBCT scanner.

Figure 4:

WBCT scans were acquired with the patients standing on the relatively small HiRise platform (A) with their feet aligned to the platform (B). This resulted in a narrow and somewhat artificially rotated double-legged stance orientation. In this orientation, images of the knees (C) and hips (D) indicated the patient’s femurs were internally rotated and adducted.

Figure 5:

Patient-specific differences in pelvis orientation from the standardized NWB-ISB model (blue). The WBCT-sagittal model (orange) differed from the NWB-ISB model by incorporation of patient-specific sagittal orientation differences of the pelvis and femur compared to the ISB-defined standing position. The WBCT-combo model (green) added to the WBCT-sagittal model by incorporating the average coronal and axial rotation from the generic motion capture data. Finally, the WBCT-original model replaced the motion capture-based coronal and axial rotations with WBCT-apparent coronal/axial rotations. Implementation of these rotational differences was performed on the pelvis component of each patient model, but are illustrated here as femur differences for easier visualization.

DEA calculation of cartilage contact stress

All four models for each hip were loaded identically with the same walking gait cycle data that was the average of gait patterns collected from 10 acetabular dysplasia patients using motion capture and musculoskeletal modeling [33]. Those hip joint reaction forces (reported in percentage bodyweight) were scaled by individual patient bodyweight, and continuous hip joint reaction force and rotation angle combinations spanning the stance phase of walking gait were discretized into 13 independent loading steps. In each model, the femur was free to translate in the superior-inferior direction according to load application, and the pelvis translated in the medial-lateral and anterior-posterior directions to achieve a seated, congruent joint position. Femur and pelvis rotations were prescribed completely based on the discretized rotation angles of the applied gait pattern [31]. According to DEA convention, bone was modeled as a rigid body [35], and cartilage was assigned homogenous isotropic linear-elastic definition (E = 8 MPa, ν = 0.42) [36, 37]. DEA computations were performed using a custom Newton’s method solver developed in MATLAB (The MathWorks, Natick, MA) [3] that was previously validated in cadaveric hips [31].

To assess regional differences in the contact mechanics calculated with the different model initialization approaches, the acetabular surface was subdivided into six subregions. Division was based on a plane defined perpendicular to a line connecting the inferior edges of the posterior and anterior acetabular lunate and passing through the centroid of a sphere fit to the acetabular lunate surface. This plane was rotated ±45 degrees around a mediolateral axis through the acetabular lunate centroid to separate anterior, superior, and posterior acetabular subregions. Each subregion was further subdivided into lateral and medial regions based on half of the distance between the lateral and medial edges of the acetabulum [31].

Statistical Analysis

The main comparisons in this study were between the NWB-ISB and our WBCT-sagittal models because this pair included true patient-specific information that was not confounded by foot position inside the WBCT scanner. 2D Pearson correlation coefficients between peak contact stress and contact area from the NWB-ISB and WBCT-sagittal models were computed in MATLAB for the entire acetabular surface, and for each of the six individual acetabular subregions, for each patient. Agreement and bias were assessed using Bland-Altman plots of peak and mean contact stress. To determine the significance of differences in computed contact mechanics between specific pairs of model initialization approaches (NWB-ISB/WBCT-sagittal, WBCT-sagittal/WBCT-combo, WBCT-sagittal/WBCT-original, WBCT-combo/WBCT-original), Wilcoxon matched-pairs signed-rank tests were performed on peak and mean contact stress and contact area in SAS 9.4 (SAS Institute Inc, Cary, NC, USA). For each Wilcoxon test, a Holm-Bonferroni multiplicity-adjusted p-value was obtained from SAS, and any p-values ≤ 0.05 were then considered statistically significant. This analysis was performed independently for the whole joint and for each of the six acetabular subregions.

Results

Patient-specific sagittal plane rotation derived from WBCT averaged 13° (range: 4° – 22°) of anterior pelvic tilt and 9° (range: 2° – 15°) of femoral extension. Pelvic axial and coronal rotations while standing were subtle, with values of 0.1° ± 0.4° and 0.1° ± 2.4°, respectively. The femur averaged 22.4° ± 8.2° of internal rotation and 3.9° ± 1.6° of adduction. Patient-specific, sagittal plane hip angles (combined pelvic and femoral rotation angles) averaged 3.8° ± 6.4° flexion. This corresponded to average sagittal plane initialization angles of 1.5° flexion with a range from 6.2° extension to 12.9° flexion after correction for generic standing (all WBCT models).

Whole-joint, point-by-point 2D Pearson correlation coefficients between WBCT-sagittal and NWB-ISB initialized models averaged 0.95 (range: 0.63 to 1.00) for peak contact stress and 0.87 (range: 0.64 to 1.00) for total contact area. Correlation coefficients were similarly positive for each patient and each subregion (Figure 6). Bland Altman plots comparing ISB-NWB and WBCT-sagittal models (Figure 7) indicated a negligible 0.02 MPa bias in mean contact stress (95% limits of agreement (LOA): −0.52 to 0.55), and peak contact stress had only one value falling slightly outside of the limits of agreement with no existing bias (LOA: −2.13 to 2.15).

Figure 6:

2D correlation coefficients for computed contact stress and the corresponding contact area between WBCT-sagittal and NWB-ISB models were calculated in each subregion. Models are sorted by increasing LCEA in the leftmost column. Correlation coefficients closer to one or negative one indicate strong positive and negative correlation, respectively, whereas values close to zero indicate weak correlation. Cells with a dash signify no contact stress was predicted in that region for either the WBCT-sagittal or NWB-ISB model throughout the entire gait cycle.

Figure 7:

Bland-Altman plots were constructed for mean and peak contact stress and show no systematic bias or patient-specificity in calculations made using the WBCT-sagittal and NWB-ISB models. The x-axis is the averaged value of the contact stress between model initialization techniques for each patient, and the y-axis is the difference in computed stresses for the associated patient. Upper and lower limits of agreement are shown by the black dashed lines, and the bias is shown by the dashed grey line. The bias is very small, as indicated by the gray dashed lines being very close to or on the zero line.

WBCT-sagittal models had an average whole-joint peak contact stress, mean contact stress, and average contact area which was not significantly different than that of the NWB-ISB models (Table 1). When averaged across patients, differences in acetabular subregion contact stress and contact area were not significantly different between the WBCT-sagittal and NWB-ISB models, however, for any given patient, there were regional differences that did not relate to severity of dysplastic deformity (Figure 8). Addition of generic axial and coronal rotation (WBCT-combo) lowered whole-joint peak contact stress (p = 0.014), but did not significantly reduce mean contact stress (p = 0.197) or increase contact area (p = 0.071) relative to the WBCT-sagittal model. Incorporation of WBCT-derived axial and coronal rotation (WBCT-original) produced more significant reductions in whole-joint peak contact stress (p = 0.001) and increases in contact area (p = 0.001) relative to the WBCT-sagittal models. These major differences were evident on the whole-joint scale, as well as in pronounced regional differences in contact area and peak contact stress (Table 1).

Table 1:

Summary table describing contact mechanics and pairwise comparisons between different modeling techniques. Statistically significant values (p ≤ 0.05) are bolded.

			DEA Models				Wilcoxon p-Values Between Models
Peak Contact Stress [MPa]			NWB-ISB	WB-Sagittal	WB-Combo	WB-Original	NWB-ISB vs. WB-Sagittal	WB-Sagittal vs. WB-Combo	WB-Sagittal vs. WB-Original	WB-Combo vs. WB-Original
	Whole Joint		15.1 (2.1)	15.1 (2.8)	13.2 (2.3)	11.9 (1.9)	0.922	0.014	0.001	0.006
	Lateral	Anterior	13.7 (2.9)	13.3 (3.0)	11.5 (2.8)	10.6 (2.7)	1.000	0.029	0.006	0.422
		Superior	9.4 (3.1)	9.2 (3.1)	9.5 (2.3)	8.9 (2.5)	1.000	0.915	1.000	0.428
		Posterior	4.1 (2.5)	4.1 (2.7)	5.1 (2.5)	5.6 (2.8)	1.000	0.824	0.352	0.264
	Medial	Anterior	10.9 (3.7)	10.8 (4.8)	9.6 (4.1)	9.1 (2.9)	1.000	0.210	0.352	1.000
		Superior	8.8 (2.7)	8.7 (2.8)	8.4 (2.9)	8.5 (2.6)	1.000	0.898	1.000	1.000
		Posterior	2.9 (2.3)	3.0 (2.1)	2.9 (3.2)	3.5 (3.1)	1.000	0.824	1.000	0.047
Contact Area [mm2]	Whole Joint		710 (114)	721 (136)	784 (117)	858 (117)	0.638	0.071	0.001	0.001
	Lateral	Anterior	188 (69)	188 (66)	185 (68)	188 (70)	1.000	1.000	0.750	0.718
		Superior	227 (55)	230 (70)	278 (62)	282 (59)	1.000	0.006	0.006	0.718
		Posterior	15 (14)	19 (22)	30 (29)	44 (39)	1.000	0.255	0.040	0.012
	Medial	Anterior	103 (66)	104 (65)	111 (73)	123 (67)	1.000	1.000	0.035	0.213
		Superior	161 (65)	164 (62)	149 (56)	179 (55)	1.000	0.960	0.730	0.096
		Posterior	16 (23)	15 (22)	30 (42)	42 (42)	1.000	1.000	0.060	0.040

Figure 8:

Analysis of computed peak contact stress and total contact area for each individual hip showed patient-specific differences between results from the standardized (NWB-ISB) model and a model constructed with WBCT-derived sagittal tilt (WBCT-sagittal). These differences were regional, and as indicated by the arrows for one selected region, could result in either higher or lower values depending on the modeling technique. Contact area is averaged throughout the stance phase of gait for each patient’s six acetabular subregions. Data are arranged by decreasing severity of hip dysplasia based on radiographic measurement of lateral center edge angle. Blue bars are NWB-ISB models and orange bars indicate the WBCT-sagittal models.

Discussion

The objective of this study was to evaluate how inclusion of patient-specific weightbearing pose information impacts DEA-computed contact stresses. The WBCT-sagittal and NWB-ISB models did not predict significantly different joint-wide contact stress or contact area, indicating that standardized model orientation methods provide a reasonable overall approximation of stress when patient-specific functional standing orientations are unknown. However, there were patient-specific local differences in contact area and contact stress between the WBCT-sagittal and NWB-ISB models (Figures 8 & 9), and when incorporating coronal and axial rotation (WBCT-combo and WBCT-original). Therefore, the ultimate application of contact stress data should be taken into consideration when determining the importance of including patient-specific weightbearing stance information in the underlying models.

Figure 9:

Contact stress difference maps at discrete time intervals of heel-strike, mid-stance, and toe-off arranged by patient-specific sagittal tilt deviations from ISB neutral. Locations with higher predicted contact stress when the model was in the WBCT-sagittal orientation are shown yellow-to-red, while locations that had higher contact stress when the model was in the NWB-ISB orientation are shown cyan-to-blue. Dysplasia severity is indicated for each case by the lateral center edge angle (LCEA) included in parentheses. For visualization purposes, all maps are shown from a right hip perspective (anterior on the right), however there were 4 left and 7 right hips in the cohort.

Other studies have utilized elements of personalized model positioning that can be derived from standing radiographs or collected patient kinematics [10, 38]. For example, Kitamura, et al. constructed a dysplastic FEA model that utilized radiographically apparent patient-specific pelvic tilt while standing [14]. Their findings generally agreed with our own in that contact area was moderately greater with addition of standing sagittal tilt compared to a standard orientation, while maximum contact pressure was nearly identical. However, their technique only incorporated pelvis orientation and omitted the functional orientation of the femur, which was found to be not negligible in our cohort (e.g. average 9° of femoral extension [range: 2° – 15°]). Aitken, et al. explored the effects of standardizing or including patient-specific femoral version in a cohort of DEA models of dysplastic hips, and found that the average computed contact stress were not significantly different with the inclusion or exclusion of patient specificity [39]. This paralleled a cadaveric study illustrating femoral rotation differences of <15° resulted in subtle movements of location, but not magnitude, of peak acetabular contact stress [17]. However, what these types of studies have in common is that differences in computed contact mechanics for any individual are lost once averaged across the study cohort, although, the individual may have substantial differences in location or magnitude (Figures 8 & 9). As local regions of high contact stress have been associated with cartilage degeneration in dysplastic hips [31], incorrectly capturing these regions by not including sufficient patient specificity significantly detracts from the ability to predict cartilage damage from computational models.

The functional orientation of two-legged stance is not a common reference frame from which to prescribe joint rotations and load application to simulate walking gait in computational models. A conversion from standing to the coordinate system in which gait is reported will always be needed. While Skalshoi et. al. found no differences in standing kinematics of dysplastic patients and controls [34], the data were analyzed in the Grood and Suntay joint coordinate system [40], and the actual values were not reported. Therefore, a conversion had to be derived for this work from motion-capture data acquired at our institution for an age-matched normal cohort. Although kinematic data collection by optical marker-based tracking can introduce soft tissue artifact [41–43], which can result in as much as 5.8° worth of error in the sagittal plane [43], our work applied the same factor to each patient model, making any error systematic across our cohort.

While WBCT is a novel technique that provides patient-specific functional initialization at the time of scan acquisition without a need for separate gait capture or analysis of standing radiographs [44], there were two key limitations regarding the use of WBCT in our model development. First, standing WBCT is a technology still being adapted for use in the hip joint. Scan parameters, including radiation dosage [45], are continually changing to reduce image noise, increase contrast between bone and soft tissue, and bring image appearance into closer agreement with that of clinical images. To prevent inconsistencies in model geometry resulting from segmenting different type of CT images from influencing differences in computed contact stress mechanics, we performed DEA using the more accurate bony geometry obtained from the non-WBCT scan and a version of that geometry that was spatially registered to the WBCT segmentation. In order to deliver on the promise of supporting development of patient-specific computational models incorporating both geometry and position, further development of WBCT acquisition/reconstruction techniques will be needed. The second major limitation of current WBCT technology for hip applications was the size of the foot platform which necessitated placing the feet closer together than is typical for 2-legged stance in most participants (Figure 4). Therefore, while the WBCT scanner was able to capture a loaded relationship between the pelvis and femur that had previously been unaccounted for in computational studies, the overall femur orientations were internally rotated and slightly adducted in comparison to normal stance outside the scanner. This oddity in stance resulted in significant differences in full patient-specific (WBCT-original) contact mechanics calculations compared to the other model initialization techniques utilized. A larger foot platform in the scanner would allow for more natural double-leg stance of a patient.

The DEA methodology utilized in this study for prediction of contact stress is limited by the simplification of bone to a rigid surface and cartilage material properties from non-linearly elastic, time-dependent, poroelastic [46] to linear elastic. DEA models also exclude the acetabular labrum, which plays a varying role in dysplastic acetabular mechanics [47]. However, even with these limitations, the accuracy of DEA-computed contact mechanics has been thoroughly validated in the hip [7, 31, 48], and it was chosen as the computational approach for this work to allow rapid execution of the large number of models needed to evaluate the effects of initial model positioning. The gait loading parameters utilized in this study were an average gait pattern of multiple individuals with hip dysplasia [33], which may not be an accurate representation of walking gait for any given patient in our study. Future studies investigating the effects of patient-specificity will need to capture patient-specific movement. However, application of the same literature-based gait information in this work allowed us to isolate the effects of initial model position for systematic study.

Conclusions

These results show that incorporation of WBCT-derived patient-specific stance into computational models of dysplastic hips produced modest differences in computed whole-joint contact stress. This suggests that the approach of using anatomic landmark-based coordinate systems is acceptable when considering peak or mean contact stress on a whole-joint scale. However, total contact area differed significantly between model initiation techniques, and inclusion of patient-specific pelvis/femur orientation from the WBCT caused variable, patient-by-patient, regional effects on the computed contact area and stress metrics. These variable mechanical differences introduce the potential to underestimate contact stress in critical weightbearing regions of the acetabulum (such as the superior lateral region) within the context of an overall reasonable approximation of whole-joint contact stress. Such localized differences could have substantial implications for focal cartilage degeneration [31]. The technological advancement provided by the emerging technology of standing WBCT presents the opportunity to increase patient specificity and thus accuracy of computed contact mechanics without the need to perform adjunctive procedures such as full kinematic and kinetic gait analysis.