To assess the accuracy of subject-specific hip model predictions of anterior femoroacetabular clearance for squatting and sitting FADIR, we compared model predictions to direct measurements of anterior femoroacetabular clearance made using open MRI scans of the hips in the same postures. These two maneuvers require extreme hip angles, which are proposed to provoke impingement. Squatting represents an active impingement provoking maneuver while sitting FADIR represents a passive impingement provoking posture.
We recruited 33 participants aged 28–56 years old, including 9 with cam and/or pincer morphology and with pain (CPM+), 13 with cam and/or pincer morphology and without pain (CPM-), and 11 controls. All participants were recruited from the original Investigation of Mobility, Physical Activity, and Knowledge Translation in Hip Pain (IMPAKT-HIP) cohort.
The IMPAKT-HIP cohort is a population-based sample of 500 Caucasian people recruited through random-digit dialing of households in greater Vancouver [29, 30]. In all participants, morphology of the femoral head-neck contour and acetabular coverage were assessed using supine Dunn view radiographs of the hip and standardized weight-bearing anteroposterior radiographs of the pelvis . Hips were identified as having cam morphology if the alpha angle  was greater than 55° on the Dunn view radiograph . Hips were identified as having pincer morphology if they had a lateral center edge (LCE) angle [34, 35] greater than 40° and/or a positive cross-over sign on the anteroposterior radiograph . Hips identified as having both cam and pincer morphologies were classified as mixed. The presence of hip pain was defined as participant-reported pain in the groin and/or upper thigh lasting for 6 weeks or more and/or for 3 or more episodes during the past 12 months. This definition was designed to exclude pain due to soft tissue injuries/deficiencies and identify pain originating only from the hip. The study hip was defined as the hip with radiographic CPM. If CPM was present in both hips, then the hip with more severe pain was defined as the study hip. If equal or no hip pain was reported, the study hip was randomly selected. Groups were defined as follows: CPM+ hips were positive for the presence of pain and had at least one radiographic CPM; CPM- hips were negative for the presence of pain and positive for at least one radiographic CPM; Controls were negative for the presence of pain and had no radiographic CPM. Recruitment and screening for the original IMPAKT-HIP study spanned 1.5 years.
For our current study, which occurred a mean (SD) of 5.7(0.5) years after the original IMPAKT contact, phone screening was completed to recruit participants. The same definitions as the original IMPAKT study were used for CPM, hip pain, and the study hip. New exclusion criteria were considered for the current study, including previous lower limb surgeries, injuries, or any neurological conditions that affected everyday recreational or sporting activities over the past 12 months, a history of any inflammatory or autoimmune diseases, avascular necrosis of the hip, planned or previous lower limb joint replacement, or physician-diagnosed lower limb joint osteoarthritis. The Clinical Research Ethics Board of the University of British Columbia approved the study, written informed consent was obtained from all participants, and the study was conducted in accordance with the Declaration of Helsinki.
We collected 3D kinematics data of each participant during squatting and sitting FADIR maneuvers at 120 Hz using a fourteen-camera motion capture system (Motion Analysis Corporation, Santa Rosa, CA). Fifty passive retroreflective markers applied to various bony landmarks were used for motion tracking. Bilateral bony landmarks included the acromioclavicular joint, posterior superior iliac spine (PSIS), iliac crest, anterior superior iliac spine (ASIS), greater trochanter, anterior thigh, lateral and medial femoral epicondyle, anterior shank, lateral and medial malleoli, posterior calcaneus, medial aspect of the head of the 1st metatarsal bone, dorsal aspect of the head of the 2nd metatarsal bone, and the lateral aspect of the head of the 5th metatarsal bone. Other bony landmarks included vertebra C7, vertebra T10, right scapula, and the sternal notch. Finally, cluster plates of 4 markers were applied bilaterally on the shanks and thighs to track the movements of their respective segments during the squatting and sitting FADIR maneuvers. A static standing trial with the same foot distance and orientation as the open MRI standing posture was then conducted. After the static standing trial was completed, the medial epicondyle and malleoli markers were removed for the squatting and sitting FADIR trials.
Participants were positioned with their feet oriented anteriorly and with the most lateral aspects of the toes 22 cm apart and their arms crossed across their chest. They squatted as deeply as they could without rotating or lifting any part of their feet (Fig. 1 (a)). Maximum squat depth was measured for each participant from the ground to the lowest part of the buttocks. Motion data for 5 trials were recorded for each participant. The toe distance restriction was applied because the final position of the squatting maneuver performed in the motion analysis lab needed to be replicated in the open MRI scanner during the scanning and was dictated by the open MRI scanner bore dimensions and the necessity of positioning the study hip at the scanner isocenter.
For sitting FADIR, the participant’s study side hip was moved into the FADIR pose using a chair designed to match the bore and chair dimensions of our open MRI scanner (Fig. 1 (b)). The horizontal distance between the mid-point of the chair seat and the most lateral margin of the chair foot holder is equal to the horizontal distance between the mid-point of the open MRI chair and open MRI wall. Participants were positioned with their study hip in the middle of the chair while their foot was secured in the chair foot holder.
The chair design allowed each participant’s hip and foot to be positioned and constrained as it was within the open MRI scanner (Fig. 1 (b)) for the same posture. The study hip was flexed and then adducted and internally rotated to the maximum limit the participant could tolerate for the duration of scanning (about 30 min) using foot displacement in the horizontal and vertical direction and then moving the knee toward the body mid-line. Motion data for 5 trials were recorded for each participant.
3D hip models
To develop 3D subject-specific hip models, we scanned each participant’s study hip in supine using a sequence optimized for hip cortical bone visualization in a 3T MRI scanner (Achieva, Philips, Eindhoven, Netherlands) (Table 1). 3T scanning of the hips was performed a mean (SD) of 4.2 (0.54) years prior to the current study and spanned 1.08 years. Subject-specific 3D models (point clouds) of the femur and acetabulum were developed by segmenting these bones manually from 3T MRI scans using Analyze software (AnalyzeDirect, Inc., Overland Park, KS, US).
Open MR imaging
We scanned each participant’s study hip in supine, standing, squatting, and sitting FADIR poses with an upright open MRI scanner (MROpen, Paramed, Genoa, Italy) to a) measure impingement directly and b) measure hip angles. To measure impingement, we acquired scans in planes parallel to the femoral neck and perpendicular to the femoral neck and femoral shaft axes (alpha plane) (Table 1) in squatting and sitting FADIR. For hip angles, we acquired images in the sagittal plane (Table 1) for supine, squatting, and sitting FADIR. We also scanned each participant’s pelvis and study side knee in supine to define a hip joint coordinate system.
We applied the following protocols for the squatting and sitting FADIR postures in the MROpen.
The MROpen bed was adjusted, and several foam pads were placed on the bed to replicate the squat depth measured in the motion analysis lab for each participant. Participants were positioned so that the top foam was touching the buttocks to prevent participant motion and associated movement artifact. Foot orientation and position from the motion analysis lab were replicated in the scanner using a pair of sandals attached to a wooden plate. Several foam pads were put around the participants’ knees to minimize movement artifact (Fig. 1 (a)).
Participants were positioned in the scanner chair with the study hip at the center of the scanner. The hip was flexed, adducted, and internally rotated to the same posture used in the motion analysis study. The foot was secured against the MROpen scanner wall using a support bar. The knee was supported with several foam pads to minimize motion artifact (Fig. 1 (b)).
Hip angle measurement from MROpen
To calculate hip joint angles from the MROpen images, we first defined hip joint coordinate systems according to the International Society of Biomechanics (ISB) recommendations  using the supine scans. To define the hip joint coordinate system, a Cartesian coordinate system was first defined for both the femur and pelvis. The pelvis coordinate system was defined using the right and left anterior superior iliac spines (ASIS) and right and left posterior superior iliac spines (PSIS). Orientations of pelvis coordinate system unit vectors were as follows: Z-axis: parallel to the line connecting the right and left ASIS pointing laterally; X-axis: parallel to a line lying in the plane defined by right and left ASIS and the midpoint of the right and left PSIS, perpendicular to the Z-axis, pointing anteriorly; Y-axis: cross product of X and Z pointing superiorly. The origin of the pelvis coordinate system was defined as the center of the best fit sphere to the acetabulum lunate surface. The femur coordinate system was defined using medial and lateral femoral epicondyles and the center of the best fit sphere to the femoral head surface. Orientations of the femur coordinate system unit vectors were as follows: y-axis: the line joining the midpoint between the medial and lateral femoral epicondyles and the center of the best fit sphere to the femoral head pointing superiorly; z-axis: the line lying in the plane defined by the center of the best fit sphere to the femoral head and the medial and lateral femoral epicondyles perpendicular to y, pointing laterally; x-axis: cross product of y and z pointing anteriorly. The orientations of the hip joint coordinate system unit vectors are as follows: e1: Z-axis of the pelvis coordinate system fixed to the pelvis (flexion/extension); e2: y-axis of the femur coordinate system fixed to the femur (internal/external rotation); e3: cross product of e1 and e2, which is the floating axis (abduction/adduction). The origin of the hip coordinate system was considered the same as the femur coordinate system origin. Bony landmarks, including right and left ASIS/ right and left PSIS, and medial/lateral femoral epicondyles were identified from the supine scans of the pelvis and knee, respectively. The femoral head surface and acetabular lunate surfaces were segmented from the supine sagittal scans of the hip.
To calculate hip joint angles from the MROpen images of the squatting and sitting FADIR poses, 3D models of the femur and acetabulum in the supine, squatting, and sitting FADIR postures were created by segmenting sagittal scans of the hip for these postures. 3D models in supine were registered to 3D models for the squatting and sitting FADIR postures using the finite iterative closest point (ICP) algorithm . Hip joint angles in each posture were calculated using the Grood and Suntay convention  applied to the hip joint coordinate system.
The primary outcome variable was the beta angle, which defines clearance between the femoral head/neck junction and the acetabular rim . The beta angle is measured on the same imaging plane as the alpha angle and is defined as the angle between a line drawn from the femoral head center to the most lateral bony margin of the acetabular rim and a second line drawn from the center of the femoral head to the starting point of deviation from sphericity in the femoral head-neck contour (Fig. 2). There is a significant association between negative beta angle and elevated acetabular rim contact pressures .
We calculated hip joint angles from the motion analysis data using the same joint coordinate system. The position of the joint coordinate system (found in supine) in the standing posture was determined by using transformation matrices found by registering 3D hip models in supine to 3D hip models in standing created from sagittal hip scans in supine and standing, respectively. The locations of the motion analysis markers in the identical (standing) posture used in the MROpen scanner were determined. Given the location of the joint coordinate system and the motion analysis markers in this reference standing posture, the location of the joint coordinate system for any subsequent frame of motion analysis data could be calculated, and joint angles could be determined using the Grood and Suntay convention .
Hip position for the squatting and sitting FADIR postures imaged in the MROpen scanner was matched to the motion analysis data by choosing the frame of motion analysis data that yielded the minimum least-squares error between hip joint angles in the MROpen and motion analysis data.
For the squatting and sitting FADIR postures, we compared the model prediction of the beta angle to the direct measurement of the beta angle on the MROpen scan. We determined the hip joint angles for the squatting posture in the MROpen scanner and then identified the frame of squatting from the motion analysis data that best matched these joint angles (i.e., with least-squares error in hip joint angles). Femur and acetabulum 3D models segmented from the 3T scans were positioned to this identified time frame of motion sequence using calculated transformation matrices from the supine posture (corresponding to the 3T image acquisition) to the relevant frame of motion analysis data. The femur’s position was adjusted to match the femur and acetabulum center-to-center distance calculated from the 3T MRI scans. The orientation of the alpha plane was determined, and a series of planes parallel to the alpha plane and spaced 2.5 mm apart were passed through the positioned 3D model of the hip to replicate MROpen slices. The intersection of these planes with the hip 3D models was found as the points within 0.5 mm of the defined planes. The beta angle was calculated for each plane. The minimum beta angle (for all planes) was found and compared to the minimum beta angle measured from the MROpen scans.
We calculated each subject-specific model accuracy as the absolute difference between the beta angle calculated from the hip model and the beta angle measured directly from the MROpen scans. Final model accuracy was described with the mean (SD) (and root mean squared) of subject-specific model accuracies.
We assessed the relationship between the least-squares error in hip angles and subject-specific model accuracy using the Pearson correlation coefficient (and intra-class correlation coefficient (ICC)) to better understand how model accuracy is related to how well the positions from the motion analysis data and the MROpen data are aligned.
To exclude data points where the accuracy would be affected by the difference in hip position between the MROpen posture and the motion analysis posture, we set a threshold for the least-squares error of hip angles, and only considered participants that were below this threshold in the final model accuracy calculation. This threshold was defined by including participants with the lowest least-squares error in hip angles until there was no statistically significant relationship between the least-squares error in hip angles and subject-specific model accuracy. For squatting, we had to include participants with less than 10% mean least-squares error in hip angles (10 participants) to reach no significant correlation (r = 0.47, p = 0.17) between the least-squares error of hip angles and subject-specific model accuracy (ICC = 0.041). For sitting FADIR, we had to include participants with less than 5% mean least-squares error in hip angles (7 participants) to reach no significant correlation (r = 0.63, p = 0.12) between the least-squares error of hip angles and subject-specific model accuracy (ICC = 0.0071) (3 participants who met the threshold criteria for the sitting FADIR were excluded from the final accuracy calculation because either they hadn’t completed the 3T MRI scanning or had very low-quality scans).
We assessed the relationship between model accuracy and participant body mass index (BMI) using the Pearson correlation (and ICC) to investigate how BMI affects the amount of skin-mounted marker movement relative to the hip bones and model accuracy.