Is micro-computed tomography useful for wear assessment of ceramic femoral heads? A preliminary evaluation of volume measurements

Introduction

Wear associated with hip components represents the main clinical problem of total hip arthroplasty (THA) treatment (1-2-3), and new techniques for more accurate measurements of wear are to be welcomed (4, 5). Currently, laboratory determination of wear rate of joint bearing constitutes an important aspect in preclinical validation of prostheses (6). Numerous measurement techniques are used to determine wear for both in vivo (i.e., radiostereometric analysis, radiography etc.) and in vitro (i.e., gravimetric method, coordinate measuring machine etc.) prosthetic hip joint components (4). In in vitro studies, wear can be measured quite accurately before, during and after the entire hip simulation. The gravimetric method, as suggested by Carmignato and coworkers (5), is the so-called gold standard practice adopted by orthopedic industry for wear evaluation of hip joint components. A microbalance is used for this procedure, and the mass loss of the specimens is measured following a procedure recommended by international guidelines (ISO 14242-2:2012). Currently the gravimetric method is the most commonly used measurement approach to measure wear in vitro (4-5-6-7).

Recently, volumetric methods such as the Coordinate Measuring Machine (CMM) and micro-computed tomography (micro-CT) techniques have been used as alternatives to overcome the limitations of the gravimetric method (5, 7-8-9). In particular, micro-CT is considered a promising technology for dimensional metrology (10, 11). In addition, geometrical approaches that use micro-CT would be the most direct ones to understand the effects of wear on contact geometry. In the case of a hip prosthesis, the effectiveness of micro-CT techniques to assess the volumetric mass loss during wear tests is not fully proven yet. Micro-CT is a nondestructive testing technique based on X-rays that gives quantitative 3-dimensional (3D) information for a whole object structure in a relatively short time (12-13-14). The most common micro-CT systems are equipped with a polyenergetic X-ray beam source (15). In these systems, low-energy photons are attenuated more rapidly than high-energy photons while they are penetrating dense materials (beam hardening effect), as reported by other authors (16-17-18-19-20-21). In addition to beam hardening, the accuracy of the images and the subsequent accuracy of micro-CT-based measurements, as reported by Kruth and coworkers (14), is influenced by the detector pixel size and by the quality of the reconstructed 2D images. In fact, both the choice of threshold value and the image-processing software techniques used to enhance sharpness or to cut image noise (e.g., speckles) directly affect the accuracy of dimensional measurements (22, 23).

The main goal of this study was to determine the appropriateness of micro-CT technique in assessing the volume of ceramic femoral heads in anticipation of a wear analysis. In particular, this study tried to determine the accuracy and precision of micro-CT volume measurements and to statistically compare the results with the gravimetric data.

Methods

An alumina (Biolox® Forte) femoral head (Ø = 28 mm) was used. For the gravimetric method, the mass of the sample was measured using a 4 decimal place digital microbalance (Denver Instruments, Bohemia, NY, USA) with a declared repeatability of 0.1 mg. The true density value, used to calculate the weight in a given volume, was measured by helium pycnometry (Quantachrome™ Ultrapycnometer; Quantachrome, Boynton Beach, FL, USA).

Afterwards, for the volumetric method, the alumina femoral head component was scanned 30 times with a Skyscan 1176 microtomographic system (Bruker Micro-CT; Bruker, Kontich, Belgium) using 2 different nominal resolutions (17.5- and 35-μm pixel size). The 2D detector resolution was set at 668 × 1,000 or 1,336 × 2,000 pixels depending on nominal resolution. In agreement with other protocols, source voltage and current were set to 90 kV and 278 μA, respectively, using a copper filter of 0.1 mm between the source and the sample to perform the scanning with only the hard spectrum of the beam. The X-ray beam and detector were rotated 360° with rotation steps of 0.4° and frame averaging of 3. The duration of each scan was nearly 60 minutes (Skyscan 1176 software; Bruker) to obtain nearly 1,500 images in 16-bit TIFF file format for each dataset.

The software NRecon (version 1.6.8.0; Bruker, Kontich, Belgium) was used to reconstruct these datasets to obtain the microtomographic sections using different specifications for beam hardening corrections – i.e., 25%, 40% and 50% (details in Fig. 1). For the resolution of 35 μm, the images were reconstructed with only the beam hardening of 40%, since this was found to be the most accurate percentage of beam hardening correction, from the analysis of datasets with a resolution of 17.5 μm. After reconstruction, 4 groups of 15 datasets were thus collected.

A new dedicated software (CTAn version 1.13.11.0; Bruker, Kontich, Belgium) was used to analyze reconstructed images. First, the density phase in the images was segmented considering a gray value threshold calculated using the method proposed by Otsu (24). This method automatically finds those thresholds that minimize interclass variance of the gray-scale histogram. The volume in cubic millimeters was directly calculated on these segmented images. Subsequently, to minimize the noise in the digital datasets, 2 levels of despeckle were applied: first all of the 2D objects except the largest one, as previously identified by the Otsu thresholding, were removed, and then the same operation was repeated in the 3D space. This procedure was also carried out on reconstructed datasets – i.e. on gray-level (nonbinarized) images after an unsharp mask filtering. This method enhances the edge details by adding the difference between the original and the blurred images to the original. With this image analysis process, 24 groups of 15 datasets each were collected (Fig. 1).

SPSS software (version 18.0; IBM Inc., Chicago, IL, USA) and OpenStat (open-source software by G.W. Miller, [email protected]) were used to perform the statistical analysis. In particular, the statistical analyses were performed in 2 steps: during the first step, 1-way ANOVA with relevant post hoc tests (Dunnett and Bonferroni) (25, 26) was used to select a reduced number of groups to compare the 2 above-mentioned methods (gravimetric vs. micro-CT). The statistical level of significance was set at a p value <0.05. During the second phase, our analyses were performed to rule out statistical differences. Thus, 1-way ANOVA with relevant post hoc tests (Dunnett and Bonferroni), Student’s t-test and Kruskal-Wallis analysis with the relevant post hoc test (pairwise) were used to find the best method equivalent to the gravimetric ones.

Results

To determine which micro-CT protocol, and consequently which group, was the most similar or equivalent to the gravimetric method in measuring volumes of alumina femoral head components, 15 repeated measurements of 24 different groups were analyzed. In the first phase of the analysis, descriptive statistics were generated (data were normally distributed), and the means with the relevant standard error (SE) and coefficient of variance (CV) of both gravimetric and experimental groups are presented in Table I.

SE and CV values demonstrated the reliability of the micro-CT analysis technique. The variability in volume measurements followed an increasing trend as the group number incremented. The greatest variability was found for groups 19 to 24 (CV = 0.23%) corresponding to the micro-CT analysis protocols with the lower resolution of 35 µm. The lowest variability was found for volume measurements of groups 1 to 6 (range of CV 0.10%-0.11%) corresponding to the micro-CT analysis protocols with the lower beam hardening correction (25%).

The first phase of the statistical analysis using 1-way ANOVA with relevant post hoc tests (Dunnett and Bonferroni) and a p value <0.05 between all 24 groups and the control group resulted in the selection of 6 groups to be statistically reassessed: groups 7 to 12 (Fig. 2). In fact, these 6 groups were different without statistical significance, compared with the control group, whereas all of the other groups were significantly different, with a p value always close to zero. The 6 nonsignificantly different groups corresponded to the micro-CT analysis protocols with a beam hardening correction of 40%.

In the second phase of the statistical analysis, the p values calculated using 1-way ANOVA with relevant post hoc tests (Dunnett and Bonferroni), Student’s t-test and Kruskal-Wallis analysis, with the relevant post hoc test (pairwise) between the 6 groups and the control group, were compared. The results of this second phase are shown in Table II. The greatest p value was found for group 10. This group corresponds to a 3D Otsu segmentation after an unsharp masking filtering of the image dataset. The dataset was obtained after a setting of 17.5 µm/pixel and a beam hardening correction of 40%. The micro-CT images of the most and least similar groups compared with the control group are shown in Figure 3.

Discussion

Currently, prosthetic failures represent a clinical problem, and eliminating or reducing wear would play a crucial role in preventing them. In this work we asked whether a new micro-CT approach would result in agreement with the so-called gold standard gravimetric method, in measuring the volume of ceramic femoral heads. The gravimetric method is the standard procedure to measure loss of mass. Among the volumetric-based methods, micro-CT has some advantageous characteristics such as the relatively short scanning time for each specimen, the ability to evaluate specimens without touching their surfaces and the capability of analyzing the specimen structures even locally. However, the use of a micro-CT-based procedure to assess the wear of ceramic femur components has not been fully proven and validated yet. For this purpose, the volume of ceramic femoral heads was calculated with a commercial high-resolution micro-CT system, which is relatively frequently suited for routine laboratory or preclinical applications. Furthermore, state-of-the-art micro-CT analysis on prosthesis component wear considers only those made of polyethylene (8, 27, 28).

After 2 subsequent statistical analyses, we identified a new protocol in which volumetric mass loss was comparable between the micro-CT and the gravimetric method. The volume measurement of the best matching group differed from the measurement of the gravimetric group by less than 1 mm³ on average. In addition, the variability in volume measurements was 0.14% for CV, with an SE of 2.95 mm³.

Despite the statistical overlap, the relative SE was still higher than that for the gravimetric group. The aim of this preliminary study was to evaluate the possibility of using a nondestructive, imaging-based method. The appreciable advantages of using the micro-CT technique in wear assessments are the possibility of conducting longitudinal quantitative analyses and localizing the wear features on the hip replacement surface. The process of wear assessment will involve measurements based on the volume difference (before and after) as a result of in vitro wear tests. From this perspective, the difference of 1 mm³ observed between the chosen experimental method and the gravimetric one may not be relevant in determining the degree of wear. As regards the SE, the issue could be more problematic. Other studies have shown that the wear rate in the first phase of in vitro testing of alumina vs. alumina is between 0.1 and 0.2 mm³ for millions of cycles (29, 30).

Among the possible causes that affect the precision of micro-CT, we can identify (i) the possibly not perfect alignment of the specimens during the acquisition, (ii) the variability of the polychromatic X-ray beam and (iii) subsequent variability in beam hardening effect. Interestingly, using beam hardening correction algorithms with increasing intensity, whose effectiveness in improving the dimensional accuracy has not yet been verified (31), an increasing variability was detected.

This study has some limitations connected mainly with the small number of femoral heads analyzed. A study of many femoral prosthetic components would allow the elimination of systematic errors. This could result in a reduction in the measured uncertainty, and the samples could be analyzed varying the orientation in the 3 spatial planes during acquisition.

Further work will apply the protocol to the unworn and worn specimens of in vitro wear simulations to evaluate mass loss. Wear analysis conducted using micro-CT will be geometrically free. This means that no unworn reference surface geometry will be coaligned to measure the wear. The difference between the worn and the unworn volume of the specimen will just be calculated after only the image analysis operations. The advantage to this will lie in the fact that without a reference surface, the measurements will be influenced only by the micro-CT protocol adopted.

Conclusions

To date, there is still no literature available concerning international standards or guidelines to evaluate the uncertainty in CT measurements. The volume evaluation proposed in this study is repeatable and operator independent. The results could lay the foundations for international standardization of volumetric analysis by micro-CT, which is currently lacking due to the numerous influencing factors and their complex interactions in determining uncertain results.

In conclusion, we found a reliable metrological protocol for the volumetric analysis of ceramic femoral head prostheses, demonstrating that the micro-CT technique could be an important tool for wear assessment.