Accuracy and precision of Volumetric Matching Micromotion Analysis (V3MA) is similar to RSA for tibial component migration in TKA

Abstract

Radiostereometric analysis (RSA) is the current gold standard to determine implant migration, but it requires bone markers and special equipment. Therefore, we developed VoluMetric Matching Micromotion Analysis (V3MA), a software program for Computed Tomography‐based radiostereometric analysis (CT‐RSA). This study aimed to determine the accuracy and precision of V3MA in vitro compared to RSA and provide a clinical proof of concept. The accuracy (RMSE (Root Mean Squared Error)) and precision (SD (standard deviation)) of V3MA were compared to RSA. A tibial component was placed in 21 different positions within a cadaveric bone to assess accuracy. For precision, a total of 20 repeated zero‐migration examinations from 4 cadaveric bones with cemented tibial components were performed. In 6 total knee arthroplasty (TKA) patients 1 to 5 year migration was measured with V3MA and RSA. V3MA accuracy ranged between 0.02 and 0.09 mm for translations and was 0.01° for internal–external rotations. For RSA, the accuracy ranged between 0.03 and 0.09 mm for translations and was 0.09° for internal‐external rotations. V3MA precision ranged between 0.01 and 0.06 mm for translations and 0.02 to 0.07° for rotations. RSA precision ranged between 0.00 and 0.06 mm for translations and 0.04 to 0.25° for rotations. V3MA was successful in 6 clinical cases and no systematic bias was present. In conclusion, the accuracy and precision of V3MA were similar to RSA. Therefore, V3MA is a promising alternative to RSA in migration measurements of tibial components in TKA.

1. INTRODUCTION

Currently, the gold standard for orthopedic implant migration measurements is radiostereometric analysis (RSA), ¹ , ² , ³ measuring implant migration with an accuracy between 0.05 mm and 0.5 mm for translations and between 0.15° and 1.15° for rotations (95% confidence interval). ⁴ , ⁵

RSA is used in studies to evaluate new implant designs and surgical techniques, ⁶ since early migration is associated with late revision. ⁷ , ⁸ , ⁹ Measurements of implant migration are therefore essential to the phased introduction of new implants. ¹⁰ RSA has some logistic disadvantages as it requires a calibration cage, a special set‐up of 2 roentgen tubes, trained radiology personnel, and the insertion of markers into the bone during surgery. In addition, these markers can easily be superimposed by the metal implant (marker occlusion). To overcome these disadvantages, computer tomography‐based RSA migration analysis (CT‐RSA) has been proposed as an alternative to measure implant migration. ¹¹ , ¹² , ¹³

The decrease in radiation dose and increase in image resolution of modern CT scanners have recently increased the use of CT‐RSA. Different methods for CT‐RSA have been introduced such as CTMA (Sectra, Linköping, Sweden), CTSA, ¹³ , and AI‐based CT‐RSA. ¹⁴ To interpret results from CT‐RSA, the accuracy and precision of CT‐RSA must be at least similar to those of RSA to detect clinically relevant migrations, such as established thresholds. ¹⁵

The accuracy, comparing measurements to a “true value”, can only be determined in vitro by applying multiple displacements to the implant. For precision, in vitro experiments are also commonly used as they allow for repeated measurements in controlled and safe conditions without exposure of patients to radiation. To account for all circumstances in clinical conditions such as patient positioning and soft tissue, a clinical experiment is necessary to validate new CT‐RSA migration measurement methods.

We developed a marker‐free, volume‐based CT‐RSA software program, called VoluMetric Matching Micromotion Analysis (V3MA). This study aimed to determine accuracy and precision of V3MA migration measurements for tibial components in total knee arthroplasty (TKA) using in vitro experiments. Secondly, our method was tested in a small clinical data set as a proof of concept. V3MA results were compared with RSA results as a reference.

2. MATERIALS AND METHODS

2.1. V3MA

V3MA is a noncommercial in‐house developed software (RSAcore, V3MA, Python v3.9.7., version date 10‐02‐2023) for migration analysis using image registration. Image registration is the spatial alignment of 2 images (reference and follow‐up CT image), to measure displacements (i.e. migration) of volumes e.g. orthopedic implants relative to the host bone. The following steps were performed (technical details in Supplementary data S1):

• 1.CT images were acquired on a 320‐row detector CT scanner (Aquilion ONE, Canon medical systems, The Netherlands) at the Leiden University Medical Center (LUMC, the Netherlands). Images were reconstructed with a bone filter without metal artifact reduction algorithms. The slice thickness was 0.5 mm, and pixel size ranged between 0.30 and 0.38 mm (Table 1).
• 2.Three masks were defined semi‐automatically in the reference CT image: align mask (including bone, implant, and soft tissue if available), tibial bone mask, and tibial component mask. Segmentation in Mimics (v23.0, Materialize, Belgium) included thresholding, region grow, and morphological operations (open and close). The 3D models of these masks (. stl files) were imported into the V3MA software and used for the image registration.
• 3.Image registration of reference and follow‐up CT images was done in V3MA using Elastix. ¹⁶ Three registrations were performed using voxel intensities within the three defined masks finding a transformation mapping from one image to another: align registration (initial alignment), bone‐bone registration (using align registration as the starting point), and implant‐implant registration (using align registration as the starting point). For all registrations, a normalized cross‐correlation metric was optimized using an adaptive stochastic gradient descent optimizer in a multi‐resolution approach. ¹³ , ¹⁶ The operator visually checked if registrations were correct (Figure 1).
• 4.Migration of the implant with respect to the bone was calculated from the implant‐implant and bone‐bone registrations. V3MA coordinate system was similar to the coordinate system of RSA, with its origin in the center of mass of the implant mask (Figure 2). Migration results were expressed as translations in mm along these axes: Tx (medial‐lateral), Ty (superior‐inferior), Tz (anterior‐posterior), and rotations in degrees around these axes: Rx (anterior‐posterior tilt), Ry (internal‐external rotation), Rz (valgus‐varus rotation). Total translation in mm (TT) and total rotation in degrees (TR) were calculated using the 3D Pythagorean theorem.

Table 1.

CT parameters of CT images used in accuracy, precision, and clinical proof of concept. All CT images had matrix size 512 × 512 (rows x columns).

CT parameter	Accuracy	Precision	Clinical proof of concept
CT scanner	Aquilion ONE	Aquilion ONE	Aquilion ONE
Helical/Volume	Volume	Volume	Helical and Volume
CT tube voltage (kVp)	120	120	120 to 135
Data collection diameter (mm)	500	500	320 to 500
CT tube current (mA)	30	30	66 to 110
Reconstruction diameter (mm)	154.296	160.156^	152.50 to 195.312
Convolution kernel	FC30†	FC30†	FC30†
Focal spot sizes (mm)	0.9\0.8	0.9\0.8	0.9\0.8
Slice thickness (mm)	0.50	0.50	0.50
Pixel spacing (mm) *	0.30	0.31^	0.30 to 0.38
Effective Dose (mSv)	0.01	0.01	0.03 to 0.09

Abbreviations: CT, Computed Tomography; kVp, kilo volt; mA, milli ampere; mm, millimeter; mSv, millisievert.

^{^} for 5/20 CT volumes the reconstruction diameter was 500 resulting in a pixel size of 0.98 mm (set 4).

^† FC30 refers to a Toshiba convolution kernel used in image reconstruction.

^*Pixel spacing is equal in both directions (square) and equals pixel size assuming the space between pixels is zero.

Figure 1.

Clinical example of visual verification of registration between the reference (green) and follow‐up (red) CT image for a high‐migrating tibial component in total knee arthroplasty. This red/green checkerboard pattern is one of the visualization methods included in the V3MA software. (A) shows an example of correct implant‐implant registration of the tibial component. (B) shows an example of correct tibial bone‐bone registration indicated by proper alignment of the red and green tibial bone parts. Because of the relatively large migration for this specific patient, it is clearly visible that the tibial bone is not aligned in image A and the tibial component is not aligned in image B, both indicated by misalignment of the red and green, and, the bone and implant parts.

Figure 2.

Coordinate system of RSA and V3MA with the center of mass of the tibial component as the origin. The z‐axis points out of page/screen. Translation along the x‐axis (Tx) is medial‐lateral, along the y‐axis (Ty) is superior‐inferior, and along the z‐axis (Tz) is anterior‐posterior. Rotations are, according to the right‐hand rule, anterior‐posterior tilt around the x‐axis (Rx), internal‐external rotation around the y‐axis (Ry), and valgus‐varus rotation around the z‐axis (Rz). Axes are defined for a right knee.

2.2. RSA

RSA‐radiographs were acquired using 2 X‐ray sources (Oldelft‐Benelux, the Netherlands) angled at 40° towards each other in combination with digital radiography detectors (Canon CDXI, 434*35 cm; Oldelft‐Benelux, the Netherlands) and a uniplanar calibration cage (Medis CarbonBox 008; Medis Specials, The Netherlands). ¹⁷ RSA was performed by an RSA expert (LK) using Model‐based RSA software (v.4.2014, RSAcore; LUMC, the Netherlands) and migration of the center of mass of the tibial component surface model was calculated between reference and follow‐up RSA‐radiographs. ¹⁸ , ¹⁹

2.3. Experiments

Cadaveric human tibia bones were prepared to accommodate tibial components. For the accuracy measurements, the uncemented tibial component could move about 5 mm within the tibial shaft. For the precision measurement, 4 other tibia cadavers were used to cement the tibial component.

2.3.1. In vitro accuracy of V3MA and RSA

Accuracy (i.e. “the closeness of agreement between measured and true quantity value” ²⁰ ) was determined for all translations and internal‐external rotation applied to a Posterior‐Stabilized tibial component (Size 5, NexGen PS, Fluted, Zimmer Biomet, Warsaw, In, USA) attached to a micromanipulator (made and calibrated by Design and Prototyping department, LUMC, the Netherlands), which could move freely within a human cadaveric tibial bone (Figure 3). True translations and rotations were provided by magnetic linear encoders (accuracy: 0.02 mm/0.01°; MRS130, IESElektronics LTD, Esenler‐Istanbul‐Turkey) displayed digitally (DC‐503, Ditron, Chengdu, Sichuan, China).

Figure 3.

(A) Set‐up to determine the accuracy. Including a tibial component attached to the micromanipulator, allowing free movement of the implant in controlled steps. The cadaveric human tibial bone was fixed to a wooden board. (B) V3MA visualization of a coronal slice with tibial component mask (red) and bone mask (green).

CT‐ and RSA‐acquisition protocols were identical: 5 translations along each axis (0.1, 0.2, 0.5, 1.0, and 1.5 mm) and 6 rotations around the superior‐inferior‐axis (−1.5, −1, −0.5, 0.5, 1 and 1.5 degrees), resulting in 21 follow‐up images. Reference imaging was performed before each set of translations (3 reference examinations) and before the positive and negative rotations separately (2 reference examination), where the micromanipulator was set to the reference position (zero). The position of the entire set‐up was slightly adjusted at random between acquisitions to mimic clinical positional variations.

2.3.2. In vitro precision of V3MA and RSA

Precision (i.e. “the degree to which repeated measurements under unchanged condition show the same results” ²⁰ ) was determined based on repeated measurements without movement of the tibial component with respect to the bone. We used 4 sets of cadaveric human bones with cemented TKA (Figure 4). 2 sets included a size 5 mobile bearing Legacy Posterior‐Stabilized tibial component (MB‐LPS, Zimmer Biomet, Warsaw, In, USA). The third set had a size 7 MB‐LPS Fluted‐Stem tibial component and the fourth set had a tibial component of unknown origin. The position of the entire set‐up was changed between acquisitions to mimic differences in patient positioning. Each set was scanned 5 times in the CT scanner resulting in 1 reference and 4 follow‐up images. The experiment was repeated for RSA after attaching bone markers to the tibial bone. Only sets with known tibial components were used for RSA, resulting in 12 migration measurements instead of 16. Note that the observations used for precision measurements were not fully independent as 1 reference examination was used to perform 4 subsequent migration measurements.

Figure 4.

(A) Set‐up to determine the precision. The tibial and femoral components were cemented in the cadaveric human bones, therefore zero migration was expected. (B) V3MA visualization of set 1 to set 4 (from left to right) with masks of the tibial component (red) and the tibial bone (green).

2.4. Clinical proof of concept

We selected 6 patients from an ongoing RSA‐study with a cemented TKA (Persona PS or NexGen LPS, Zimmer Biomet, Warsaw, IN, USA) who also had 1 and 5‐year postoperative CT scans made according to standard knee CT‐protocol in our hospital. ²¹ Patient selection was based on CT pixel size below 0.40 mm to match the resolution of the previous experiments. Tibial implant migration was measured with both V3MA and RSA between 1 (reference) and 5 (follow‐up) years postoperatively. The masks used in V3MA were defined by two observers (NL and BK) to determine in vivo interobserver variability of the segmentation method.

2.5. Outcome measures and statistics

In each experiment migration was measured with V3MA and RSA (8 output variables: Tx, Ty, Tz, Rx, Ry, Rz, TT, TR). Data analysis was conducted using R (version 3.6.1, R Foundation for Statistical Computing, Vienna, Austria) and Rstudio (version 1.1.456).

2.5.1. Accuracy

Accuracy was evaluated by calculating the root mean squared error (RMSE), since the difference between measurements could be both positive and negative. ²² The error (Me‐Mt) is the difference between the estimated migration result (Me) and the true displacement imposed by the micromanipulator (Mt), which was calculated for each measurement.

2.5.2. Precision

Precision was presented with the mean and standard deviation (SD) of repeated measurements. ²⁰

2.5.3. Clinical proof of concept

Bland‐Altman plots were constructed to determine the mean difference and limits of agreement (mean ± 1.96*SD) between V3MA (NL) and RSA, and between two observers operating with V3MA (NL and BK). ²³ , ²⁴ , ²⁵ Additionally, the interclass correlation coefficient (ICC) was calculated.

2.5.4. Radiation

The effective dose (ED) was calculated by multiplying the CT scan dose length product (DLP) by the conversion coefficient (k). The k value for a CT scan of the knee is 0.0004 (mSv/(mGy*cm)). ²⁶

2.6. Ethics

Anatomical specimens were obtained from donated human bodies (Dutch Burial and Cremation Act) from the Department of Anatomy and Embryology at the LUMC. The local scientific committees of the departments of radiology and orthopedics approved the in vitro experiments.

The patients included in the clinical study gave written informed consent for the initial RSA study including CT acquisition 1‐year postoperatively and the addendum to acquire CT images 5‐years postoperatively. The study protocol was approved by the institutional review board of LUMC under number: P13.277 and registered in ClinicalTrials. gov under NCT02269254.

3. RESULTS

All registrations and mask creations performed as described in step 3 in the methods sections were correct and no reregistration or mask adjustments were necessary.

3.1. Accuracy of V3MA and RSA

V3MA accuracy (RMSE) varied between 0.02 mm and 0.09 mm for translations and was 0.01° for internal‐external rotation (Table 2), whereas RSA accuracy varied between 0.03 mm and 0.09 mm for translations and was 0.09° for internal‐external rotation (Table 3).

Table 2.

Accuracy of V3MA for imposed translations along the x‐, y‐ and z‐axis and rotations around the y‐axis (n = 21).

	Translations (mm)				Rotations (°)
V3MA Accuracy	Tx Medial‐lateral	Ty Superior‐inferior	Tz Anterior‐posterior	TT*	Rx Anterior‐posterior	Ry Internal‐external	Rz Valgus‐ varus	TR*
RMSE	0.09	0.06	0.02	0.11	0.02	0.01	0.01	0.11
Mean	0.00	−0.03	0.01	0.08	0.01	0.01	0.00	0.08
SD	0.09	0.05	0.02	0.08	0.02	0.01	0.01	0.08
Minimum	−0.21	−0.12	−0.02	0.01	−0.01	−0.01	−0.01	0.01
Maximum	0.26	0.02	0.05	0.29	0.06	0.03	0.02	0.30

Abbreviations: RMSE, root mean squared error; SD, standard deviation.

^*Micromanipulator only rotates around y‐axis.

Table 3.

Accuracy of RSA for imposed translations along the x‐, y‐ and z‐axis and rotations around the y‐axis (n = 21).

	Translations (mm)				Rotations (°)
RSAAccuracy	Tx Medial‐lateral	Ty Superior‐inferior	Tz Anterior‐posterior	TT*	Rx Anterior‐posterior	Ry Internal‐external	Rz Valgus‐ varus	TR*
RMSE	0.09	0.03	0.07	0.12	0.12	0.09	0.02	0.15
Mean	0.00	−0.01	−0.04	0.10	−0.04	−0.01	0.00	0.14
SD	0.1	0.03	0.06	0.07	0.11	0.09	0.02	0.05
Minimum	−0.21	−0.05	−0.18	0.03	−0.22	−0.17	−0.04	0.04
Maximum	0.28	0.05	0.03	0.28	0.17	0.13	0.04	0.24

Abbreviations: RMSE, root mean squared error; SD, standard deviation.

^*Micromanipulator only rotates around y‐axis.

3.2. Precision of V3MA and RSA

V3MA precision (SD) in translations varied between 0.01 and 0.06 mm and rotations 0.02 to 0.07° (Table 4). RSA precision in translations varied between 0.01 and 0.05 mm and rotations 0.03 to 0.21° (Table 5).

Table 4.

Precision of V3MA for repeated measurements (n = 16).

	Translations (mm)				Rotations (°)
V3MA	Tx Medial‐lateral	Ty Superior‐inferior	Tz Anterior‐posterior	TT*	Rx Anterior‐posterior	Ry Internal‐external	Rz Valgus‐ varus	TR*
Mean	0.00	−0.02	0.00	0.05	0.01	−0.01	0.00	0.05
SD	0.01	0.06	0.02	0.04	0.02	0.07	0.02	0.05
Minimum	−0.03	−0.14	−0.03	0.01	−0.02	−0.18	−0.03	0.01
Maximum	0.01	0.06	0.05	0.14	0.06	0.11	0.04	0.18

Abbreviations: SD, standard deviation; TR, Total rotation; TT, Total translation.

Table 5.

Precision of RSA for repeated measurements (n = 12).

	Translations (mm)				Rotations (°)
RSA	Tx Medial‐lateral	Ty Superior‐inferior	Tz Anterior‐posterior	TT*	Rx Anterior‐posterior	Ry Internal‐external	Rz Valgus‐ varus	TR*
Mean	0.00	0.00	0.00	0.05	0.08	−0.04	−0.01	0.20
SD	0.03	0.01	0.06	0.04	0.10	0.25	0.04	0.20
Minimum	−0.04	−0.02	−0.08	0.01	−0.02	−0.69	−0.09	0.03
Maximum	0.05	0.03	0.15	0.15	0.32	0.33	0.07	0.77

Abbreviations: SD, standard deviation; TR, total rotation; TT, total translation.

3.3. Clinical proof of concept

V3MA could be used in all CT images of the total knee prostheses of the 6 patients (see Table 6 in supplementary data S2 for details of CT parameters). The mean differences (limits of agreement) between V3MA (NL) and RSA were −0.05 mm (−0.49 to 0.39 mm) and −0.14° (−0.79 to 0.51°) for TT and TR, respectively (Figure 5), suggesting no significant bias. The limits of agreement did not exceed ±0.46 mm and ±0.60° for translations and rotations, respectively.

Figure 5.

Bland‐Altman plots of 6 patients (black dots) for comparison between two methods: V3MA (NL) and RSA, with the mean difference (black solid line) and the limits of agreement (red dashed lines).

The interobserver variability for the V3MA measurements was ICC: 0.8 (95%‐CI 0.6‐0.9). The limits of agreement between the two observers did not exceed ±0.25 mm and ±0.45° for individual translations and rotations, respectively (Figure 6). One outlier was observed in anterior‐posterior tilt (absolute difference: 0.67°). The mean effective dose was 0.05 mSv (range 0.03–0.09 mSv).

Figure 6.

Bland‐Altman plots of 6 patients (black dots) for comparison between two observers: V3MA (NL) and V3MA (BK), with the mean difference (black solid line) and the limits of agreement (red dashed lines).

4. DISCUSSION

We developed and validated a novel, marker‐free, CT‐based, Volumetric Matching Migration Analysis (V3MA) tool to evaluate tibial component migration. The accuracy and precision of V3MA in tibial component migration were similar to RSA. The V3MA software is available to other research groups.

The in vitro accuracy (RMSE) of V3MA migration measurements was similar to that of RSA and was within 0.09 mm and 0.01°. RSA results for precision and accuracy in our experiments were in line with the literature. ¹⁸ In the accuracy experiment, one outlier was present. For that matter 0.12 mm difference was measured for superior‐inferior translation between micromanipulator (1.50 mm) and V3MA (1.38 mm), in contrast to RSA (1.55 mm). This may be due to the shape of the implant since tibia plateau components contain a flat surface that is parallel to the CT slices. The superior‐inferior translation is also the “out of image plane” direction for CT which might also result in the outlier.

The precision of V3MA was at least as good as RSA for all translations and rotations. For superior‐inferior translations, a slightly lower precision (SD = 0.06 mm) was observed compared to other directions. However, this is still good enough for clinical use and comparable to RSA precision. The relatively lower superior‐inferior precision for V3MA could be explained by the slice thickness of 0.5 mm in contrast to a pixel size of 0.3 mm. Although 1 set of CT images had a different pixel size (0.98 mm instead of 0.3 mm) due to a larger reconstruction field of view, reanalysis of V3MA precision without this set still gave similar results. Thus in this small sample size, the difference in pixel size appeared not to result in a clinically relevant better precision of V3MA.

In the clinical setting, there was no clinically relevant bias observed between V3MA and RSA. The mean differences in implant migration measured with V3MA and RSA were close to zero and the limits of agreement did not exceed ±0.46 mm and ±0.60° for translations and rotations, respectively. These values seem high, however, one should note that V3MA is compared to model‐based RSA which also has lower precision compared to RSA. ²⁷ The difference between the methods comes from a combination of the measurement errors of both methods. Our clinical results are in line with recent literature. Angelomenos et al. compared CTMA with RSA for hip cups and found limits of agreements that did not exceed ±0.4 mm for translations and ±0.5 degrees for rotations. ²⁸ Christensson et al. compared AI‐based CT‐RSA with model‐based RSA for hip implants and found limits of agreements that did not exceed ±0.5 mm for translations and ±1.4° for rotations. ¹⁴

The largest absolute difference between observers was 0.67° in anterior‐posterior tilt. The ICC of 0.8 (95%‐CI 0.6‐0.9) for V3MA was moderate to excellent, although these clinical results are based on only 6 TKA patients all with high‐resolution CT images.

The accuracy and precision of V3MA are consistent with previously published results on CT‐RSA. ¹¹ , ¹³ , ²⁹ , ³⁰ , ³¹ Although comparisons are complicated by the use of different types of implants and different CT‐RSA methods. Scheerlink et al. tested CT‐based spatial analysis of migration in femoral hip implants, with limits of agreement of accuracy ≤0.28 mm and ≤0.20° for translations and rotations respectively, ¹³ which are larger than our results (≤0.11 mm and ≤0.03°). Brodén et al. reported accuracy (RMSE when divided t*RMSE by 2.57) of a 3D CT method in acetabular components between 0.03 and 0.12 mm for translations and 0.08–0.32° for rotations. ¹¹ For a humeral shoulder component, using CT motion analysis software (CTMA, Sectra, Linköping, Sweden), accuracy was between 0.03 and 0.11 mm for translations and 0.11–0.35° for rotations. ³⁰ We obtained accuracy values of 0.02‐0.09 mm for translations and 0.01–0.02° for rotations, suggesting that V3MA might be more accurate despite differences in implants used.

The precision (t*SD) of the 3D CT method for acetabular cups was between 0.01 and 0.09 mm for translations and 0.06–0.21° for rotations. ¹¹ The precision of CTMA for the humeral component was between 0.08 and 0.15 mm for translations and 0.23–0.54° for rotations. ³⁰ These results appear similar to V3MA precision (2.131*SD): 0.02–0.12 mm for translations and 0.04–0.14° for rotations, despite differences in methods and implants used.

One more recent study concluded that CT‐based migration analysis (CTMA, Sectra, Linköping, Sweden) is more precise than RSA in tibial component in a porcine cadaver. ³¹ They presented 95%‐CI in superior‐inferior direction of −0.06–0.10 mm and −0.09‐0.09°.

The clinical precision of tibial component migration using CT‐based migration analysis remains to be determined in future studies. One study reported clinical precision of CTMA for acetabular cups between 0.07 and 0.31 mm for translations and 0.20–0.39° for rotations. ²⁹ Precision of in vitro experimental settings is 2 to 3 times better compared to clinical precision, as has been reported previously. ¹⁹

4.1. Comparison V3MA with model‐based RSA

V3MA and model‐based RSA analysis are comparable in aspects of usability, time, and learning curve, but image acquisition of V3MA is less complex compared to RSA as it can be used with standard hospital CT‐protocols. In addition, V3MA is truly 3D since it is based on 3D image matching of truly 3D CT image data. Where 3D image segmentation is required only to distinguish bone from prosthesis, but underlying voxels are used to calculate migration. On the contrary, in model‐based RSA 3D models are projected and matched on 2D calibrated roentgen images, making the migration results sensitive to the accuracy of the models.

While CT‐RSA is based on standard CT imaging, RSA is based on specially‐calibrated stereo roentgen imaging and the disadvantage of the need for invasive marker placement in the bone. However, CT‐RSA has the disadvantage of a higher radiation dose as compared to RSA. According to the literature, a knee CT scan has an effective dose (ED) of 0.16 mSv, while paired RSA images only have 0.003 mSv. ²² , ²⁶ In our study, the maximum ED for knee CT was 0.09 mSv. This could be classified as a low‐dose CT. Although there is no clear definition of “low dose”, CT scans below 1 mSv might be considered low‐dose CT. ³² A total ED between 0.5 and 10 mSv is considered acceptable in clinical category IIa studies for adults above 50 years if it “increases in knowledge leading to health benefit”. ³³ A typical implant migration study lasts between 2 and 5 years, with approximately 5 follow‐up moments (5 CT‐scans). The ED of a clinical knee CT is sufficiently low, to allow for a multitude of scans to be made before approaching the radiation limit (100 CT scans).

4.2. Strengths and limitations

A strong point of this validation study is that the accuracy, precision, and clinical proof of concept of V3MA were tested and compared to RSA. Nevertheless, some limitations exist.

First, V3MA was only tested in one CT scanner using our hospital's standard knee CT protocol. For that matter, results may differ for different CT resolutions. ¹³

Second, only tibial components were used in the validation studies. Tibial components have a well‐defined 3D shape with sharp edges compared to other components with a more rounded shape (e.g. femoral cups, stems, femoral TKA component).

Third, in the experimental study, only internal‐external rotations were imposed for accuracy assessments. Therefore, accuracy results were not representative of anterior‐posterior and valgus‐varus rotations. In addition, Maximal Total Point Motion (MTPM), which is frequently used as an important outcome of tibial knee implants, cannot be determined in the current version of V3MA. This will be available in future versions of the V3MA software.

Fourth, the interobserver variability was determined between two operators with excellent knowledge of segmentation and V3MA use, whether these results can be extrapolated to less experienced operators, remains to be evaluated.

Fifth, conclusions are drawn from small sample sizes, and sample size calculation was not performed before the experiments. We based the sample size on former CT‐RSA studies and what was practically feasible. With the current sample size, V3MA shows similar precision and accuracy in several parameters, which are likely to be consistent in a larger sample size.

Sixth, most accuracy studies use a reference standard or ‘true value’ which is at least an order of magnitude higher than expected measured accuracy. In the current study, the accuracy of the micromanipulator measurement (accuracy = 0.02 mm/0.01 degrees) was in the same order of magnitude as our measured accuracy of V3MA. However, the relatively lower accuracy of the encoders can only negatively limit the measured accuracy of V3MA and therefore does not affect our conclusions.

Seventh, as with all CT‐RSA methods, an objective way to quantify the success of the registration, such as the “mean error of rigid body fitting” in RSA, is currently not yet available in V3MA. We are working on a quality metric of measurements.

Lastly, V3MA assumes rigid body displacement of both implant and bone. Over time, osteophytes, bone remodeling, and stress shielding resulting in relative osteoporosis could change the rigidness of bone in follow‐up imaging. Whether this interferes with the registration of V3MA remains to be determined in future long‐term studies.

5. CONCLUSION

In conclusion, our data suggest that V3MA and RSA were similar for measuring tibial TKA component migration regarding accuracy and precision. V3MA seems to be a potential alternative tool for implant migration measurements, to be used in any hospital with a CT scanner. However, larger clinical studies and using V3MA on different joint implants and designs are needed before the general use of V3MA in clinical studies.

AUTHOR CONTRIBUTIONS

N.L. processed the experimental data, performed image and data analysis, and drafted the manuscript. L.K., B.S., and B.K. conducted the experiments. L.K. also supervised the work and performed RSA analysis. R.N. and B.K. conceptualized the project. B.K. also developed V3MA software and performed clinical image analysis. All authors discussed the results and contributed to the final manuscript.

Supporting information

Supporting information.

de Laat NN, Koster LA, Stoel BC, Nelissen RGHH, Kaptein BL. Accuracy and precision of Volumetric Matching Micromotion Analysis (V3MA) is similar to RSA for tibial component migration in TKA. J Orthop Res. 2025;43:311‐321. 10.1002/jor.25989