The study used a phantom set-up for method development and evaluated the CMC and hybrid models in a female patient (65 years old, with body mass index 32.6). The patient was recruited from a randomised clinical trial (Clinical Trial NCT02301182) and had consented orally and in writing to study participation. The Helsinki II declaration was followed [12].
Implants and surgery
The Anatomic Dual Mobility Restoration acetabular system (Stryker, Warsaw, Mazovia, Poland) with a mobile liner made of X3 highly cross-linked PE (Stryker, Warsaw, Mazovia, Poland) and a ceramic size 28-mm femoral head was used in both the phantom and the patient. The hip stems were a Bi-Metric size 7 (Biomet, Warsaw, IN, USA) in the phantom and Accolade II (Stryker, Warsaw, Mazovia, Poland) size 4 in the patient. Cup/liner 56-mm size was used in the phantom and 50-mm size was used in the patient. An experienced hip surgeon inserted the components into the Sawbone hip (No 1301-165-1, Sawbones, WA, USA) and also the patient by use of a posterolateral approach.
Insertion of markers in the polyethylene
Twelve 1-mm tantalum markers (X-medics, Frederiksberg, Denmark) were placed centralised in the PE wall of the mobile liners in four groups of three markers by use of a custom-made drill guide, specific for each liner size (Figs. 1 and 2). In three of the four marker groups, one specific marker was placed 1.5 mm deeper than the other two markers, which provided a recognisable and unique pattern for each marker group.
Radiostereometric recordings
The RSA recordings were obtained using the AdoraRSA Suite (Nordic X-ray Technique, Hasselager, Aarhus, Denmark) consisting of two ceiling fixed x-ray tubes angled 40° on each other. For static RSA, two static digital detectors (CXDI-70C, Canon, Tokyo, Japan) were mounted below a calibration box (cb24, Medis Specials, Leiden, The Netherlands) for direct anterior/posterior recording (Fig. 3a).
For dynamic RSA, two dynamic digital detectors (CXDI-50RF, Canon, Tokyo, Japan) were mounted below a calibration box (cb14, Medis Specials, Leiden, The Netherlands) and recorded five images per second. From the projection of the calibration box markers, the foci position can be calculated, which enables projection of the hip implant and markers and comparison of implant component positions. The set-up allowed for a 45-degree angle on the hip joint in the cranial/caudal and anterior/posterior x-ray direction for optimal view (less marker occlusion) of the PE liner in the dual mobility cup (Fig. 3b). Soft tissue equivalence in terms of a 10-cm polymethyl methacrylate plate was placed in the recording area of the hip phantom during dynamic RSA recordings [13]. The hip was flexed to 45° and kept there while being moved in abduction/external rotation and adduction/internal rotation. This modified FADIR/FABER movement was performed passively to end-range position (Fig. 4). The dynamic RSA recordings were captured using 140 kVp and 8 mAs for the phantom and 130 kV and 8 mAs for the patient.
Combined marker configuration model
Marker configuration model-based RSA [8] requires a marker configuration model that describes the positions of the markers in the rigid body relative to each other. By fitting this model to its projections in the RSA radiographs, the position and orientation of the model are calculated, similar to model-based RSA [8, 14]. Such a model is created from one RSA frame, in which all markers are visible in both projections, using conventional RSA [8]. The method handles the occluded marker problem, but requires that all the model markers are projected on both RSA images in one RSA frame [8]. The combined marker configuration model (CMC model) builds on the same principles, but combines the marker positions and the position of the femoral head from more than one RSA frame in the model.
Combining two or more RSA frames requires at least three overlapping markers in the image pairs. By using the femoral head as one common marker in the marker model, the minimum number of overlapping markers needed for combining RSA frames is reduced to two.
For the phantom, only dynamic RSA frames were used to generate the CMC model. For the patient, dynamic hip RSA frames were combined with standard supine static hip RSA recordings to generate a CMC model with a sufficient number of representative markers.
For creating the CMC model, the detected markers of all frames were aligned using the migration function of the mbRSA software (version 4.2, RSAcore, Leiden, The Netherlands). For both the phantom and the patient, the three-dimensional marker coordinates were exported and the mean marker positions were calculated using a custom-made program in MatLab (version 2019b, The MathWorks Inc, Natick, MA). To evaluate the dispersion of markers contributing to the CMC model, we used the standard deviation of the aligned marker positions [15].
Hybrid marker model
A hybrid model was created by combining the marker registrations in the mean CMC model and the theoretic marker positions known from the computer-assisted drawings of the drill guide. The theoretic marker positions were aligned with the mean CMC model to add more information to the model and to be able to detect the specific four marker groups in the liner for precise registration of liner rotation. The hybrid model was used for detecting liner movement in static RSA follow-ups over time, where the liner rotation could be very different from one RSA recording to the next.
Coordinate systems
To define the local coordinate system of the CMC model, a base plane was fitted through the markers in the liner and the local coordinate system was redefined with the femoral head as origin and the y-axis perpendicular to the base plane of the liner. The coordinate system for the theoretic markers was created in a similar fashion, but the base plane excluded the three markers that were deeper in the liner wall. The hybrid model inherited the theoretic coordinate system. For the outer metal cup, a similar local coordinate system was defined with the origin in the centre of rotation of the cup and the y-axis (acetabular axis) perpendicular to the base plane of the cup. Lastly, the femoral neck coordinate system was defined with the femoral head as the centre and the y-axis aligned with the neck. This aligned the origins of the CMC model, the femoral head and the cup coordinate systems. In the “neutral” orientation, also the main y-axis of all objects was aligned. Therefore, all movements in the cup-liner-neck complex could be expressed by the angle between, e.g., the cup y-axis and the liner y-axis (Fig. 5).
RSA and data postprocessing
The CMC model was then fitted to the dynamic RSA recording, frame by frame, using mbRSA (version 4.2, RSAcore, Leiden, The Netherlands). The cup, liner- and femoral neck orientations were imported from mbRSA to a custom-made program in Python 3 [16]. To remove the patient movements during RSA recording, the raw-data orientation was standardised using the cup orientation relative to the calibration box from the first frame. This resulted in a constant cup orientation during the whole movement. The liner rotation was set to zero for the first frame of the recording. Orientation was described as inclination, anteversion and rotation in a radiographic coordinate system as described by Murray [17]. The radiographic inclination was defined as the angle between the longitudinal axis and the acetabular axis when projected on the coronal plane. Likewise, the radiographic anteversion was defined as the angle between the acetabular axis and the coronal plane [17] (Fig. 6). Stem angles were likewise calculated as standardised radiographic inclination and anteversion. Furthermore, the angle between the neck and liner normal (y-axis) was calculated (Fig. 5). This angle served as an indicator of contact between the neck and the liner. With contact between neck and liner, the liner should rotate relative to the cup. Movements were graphically displayed using Stata/IC (version: 16.0, StataCorp, College Station, TX, USA).