Assessment of the efficiency of a pre- versus post-acquisition metal artifact reduction algorithm in the presence of 3 different dental implant materials using multiple CBCT settings: An in vitro study

Imaging has become essential for implant dentistry planning and follow-up to monitor osseointegration and the outcome of the bone integration process. Cone-beam computed tomography (CBCT) is currently one of the most commonly used techniques for the follow-up of bone integration around the dental implant due to its ability to provide precise 3-dimensional images.1 These images have brought tremendous benefits in some dental specialties, such as periodontology, dental implantology, and oral and maxillofacial surgery.2, 3 One of the most important limitations of CBCT, however, is its susceptibility to artifact generation.4 Metallic objects significantly deteriorate CBCT image quality due to scattering, beam hardening, and streak artifacts that hinder the proper visualization of peri-implant bone and osseointegration,5 thereby making the diagnosis ineffectual and timeconsuming.6

Significant efforts have been made to develop metal artifact reduction (MAR) algorithms to reduce beam hardening effects and improve CBCT scan quality. Some CBCT machines have the option of applying a pre- or post-acquisition MAR algorithm, which is a newly introduced type of software that can reduce artifacts and enhance the quality of images produced. Pre- and post-acquisition MAR is based on standardizing voxel values by enhancing the reconstruction of the image and improving the contrast-to-noise ratio (CNR) to diminish the deleterious impacts of artifacts.7 The pre-acquisition MAR algorithm identifies the metal area in the basis projections. The metal projection data are then processed with the interpolation algorithm. Finally, CT axial images are reconstructed from the pre-processed data sets, and the metal section is recovered.8 In the post-acquisition algorithm, metal regions in each projection are segmented and modified. The reconstruction of the final CT image with the modified data is displayed. Nevertheless, due to the proximity of the voxel values of bone and metal, it is difficult to develop an algorithm for accurate metal segmentation.8 Additionally, different peak tube potentials (peak kilovoltage; kVp) can be applied to reduce metal artifacts.9

Various materials, such as titanium-zirconium (Ti-Zr), titanium (Ti), and zirconium dioxide (ZrO₂), are increasingly being used to manufacture implants with enhanced mechanical properties. For decades, Ti has been used to manufacture dental implants due to its ideal properties and favorable ability to osseointegrate.10 ZrO₂ then emerged as an alternative to overcome some limitations of Ti, such as an esthetically unfavorable appearance and artifact generation,11 and the overall survival rate of zirconia implants has been calculated as 92% (95% CI, 87–95) after 1 year of function.12 Alloying elements can be also added to Ti to improve its characteristics.13 Ti-Zr implants were recently introduced for clinical applications, and the mean survival and success rates of Ti-Zr implants have been reported as 98.4% and 97.8% at 1 year after implant placement and 97.7% and 97.3% at 2 years, respectively.14 To date, insufficient information has been made available to compare artifacts due to these 3 different types of implants in CBCT scans using pre- and post-acquisition MAR, as compared to images without MAR. Therefore, it is necessary to investigate the difference in artifacts induced by Ti-Zr, Ti, and ZrO₂ implants in CBCT scans.

The objective of this study was thus to evaluate and compare the effects of pre- and post-acquisition MAR on artifacts produced by 3 common types of implant materials at 2 different kVp settings, to determine whether they have the intended effect (artifact reduction relative to images taken without the MAR algorithm) and whether any method is better than another for any of the implant metals.

The intensity of artifacts generated when there was no metal implant material was measured to establish a baseline for comparison with the artifacts generated by implant materials. With no MAR setting, the intensity of the artifacts generated under the conditions of this experiment was 168 at 84 kVp and 172 at 90 kVp. The values for the artifacts using post-acquisition MAR were essentially the same: 169 at 84 kVp and 172 at 90 kVp. For pre-acquisition MAR, slightly higher values were observed (221 at 84 kVp and 200 at 90 kVp). Although the intensity of the artifacts was measured 5 times for each condition, there was no variation in the artifact intensity measurement within any of the 6 conditions.

When the implant materials were imaged, the mean value of artifact intensity with no MAR (across all 6 combinations of materials and kVp settings) was 1005, while the mean value of artifact intensity with either setting of MAR was 268, reflecting a statistically significant difference (P<0.05) (Fig. 2). The difference in mean values can readily be seen in the boxplots in Figure 3. To confirm the efficacy of the MAR algorithms, the data were divided into 2 sets: images obtained with no MAR, versus images obtained with either of the 2 MAR settings. Boxplots of the artifact data from the no-MAR images are shown in Figure 4 to assist in visualizing how the implant material affected artifacts. The Ti-Zr images had the highest median level of artifacts, but the least amount of variability, whereas the images with ZrO₂ as the implant material had the lowest median amount of artifacts, but an extreme amount of variability. The Ti material had moderate levels of both artifact intensity and variability.

Fig. 2
Example of axial images shows the efficacy of the metal artifact reduction (MAR) algorithm. A. 84 kVp and no MAR. B. 84 kVp and pre-acquisition MAR.

• Click for larger image

Fig. 3
Box plots of artifacts associated with metal artifact reduction (MAR) settings, showing the aggregation across all materials and kVp settings. The whiskers of each boxplot show the minimum and maximum values within each data set, the thick middle line is the median value, and the bottom and top of the boxes are the first and third quartiles, which indicate where the middle 50% of the data lie.

• Click for larger image

Fig. 4
Box plots of artifacts by implant material aggregated across all metal artifact reduction (MAR) and kVp settings.

• Click for larger image

With the data from no-MAR setting removed, the 3-factor ANOVA indicated no detectable difference in artifacts between the 2 remaining MAR settings (P=0.25), but a statistically significant difference in the artifacts generated between at least 2 implant materials (P<0.05) and between the 2 kVp settings (P<0.05). No interaction term was significant. The 90 kVp setting generated less intense artifacts on average than the 84 kVp setting. The Tukey HSD test indicated that with the MAR setting on, the Ti-Zr alloy generated significantly less intense artifacts than the ZrO₂ alloy (P<0.05). The mean intensity of artifacts generated by Ti fell between the other 2 implant materials, but was not statistically different from either.

Figure 5 illustrates the distribution of artifact intensity across the 6 combinations of implant materials and kVp settings, aggregated across the MAR setting, to assist in visualizing the artifacts for each modality. This demonstrates that the 90 kVp setting resulted in a less intense artifacts and that Ti-Zr generated less intense artifacts than ZrO₂. These findings also suggest that artifacts can be minimized by using the Ti-Zr alloy at the 90 kVp setting, with either MAR setting.

Fig. 5
Box plots of artifacts by implant material and kVp setting aggregated across the pre- and post-acquisition MAR settings

• Click for larger image

Conventional CT has limited value compared to CBCT for assessing the peri-implant bone tissue, and there has been an increasing interest in applying CBCT to evaluate osseointegration around implants.2 However, the main shortcoming of CBCT is the appearance of dark areas around high-density materials that cause less reliable interpretation of images. Therefore, MAR algorithms have been added into CBCT machines to correct artifacts.

According to the results of previous studies, the MAR algorithm increases the CNR and decreases the noise when there is a high-density material within the FOV. However, Vasconcelos et al. reported that the CNR was systematically better when they did not use the MAR algorithm for ZrO₂ implants.3 Similar results were shown by Bezerra et al.,15 who reported that CNR was reduced by applying the MAR algorithm when there were metal posts within the FOV. In addition, some other studies reported that the MAR tool may neither improve the voxel values nor eliminate artifacts entirely.3 Therefore, MAR may not generate a more precise scan. Nevertheless, Bechara et al. showed that applying MAR resulted in a higher CNR and a significant change in mean voxel levels when metal was present in the FOV.9 These findings are in concordance with the findings of Demirturk Kocasarac et al.16 Those authors used the MAR algorithm in the presence of 4 different types of root end filling materials and showed that MAR markedly minimized the metal artifacts and increased the CNR, suggesting that it improved image quality.16

Possible reasons for the contrasting findings among the abovementioned studies may be that some of the previous authors used the CNR to evaluate image changes. Instead, this study used a line profile measurement in the area of interest next to the implant. The line profile provides a linear measurement of the difference between the highest and lowest gray values, and not overall image quality. The CNR is one of the main factors influencing the image quality in CBCT, but was not within the scope of this study. In addition, decreased artifact values are seen when MAR is used, but how the images might look without the implant material in the FOV is unknown. Therefore, it is unclear whether the parameters used or MAR algorithm application actually restore the image content to offer a closer demonstration of the underlying reality (e.g., whether the adjusted image represents the actual object attenuation better than the non-adjusted image, or whether these settings are all merely a matter of iteration or interpolation aiming to decreasing the influence of the high-attenuating materials). The gray values attained by using the MAR algorithm might be different from those that would be computed if the same image were obtained with no metal in the FOV.7 Nonetheless, besides all its advantages, the reconstruction time is longer when the MAR algorithm is applied.17

Bechara et al.17 found that the quality of scans was better when applying pre-acquisition MAR, rather than postacquisition MAR, when they evaluated the effect of the MAR algorithm in the proximity of a metallic bead. In contrast, Bechara et al.6 suggested using post-processing MAR when they investigated the effect of the MAR algorithm on the detection of root fractures. In another study done using the Planmeca ProMax device, Kamburoglu et al.18 did not find a statistically significant difference in the discernment of peri-implant bone defects with and without use of the MAR algorithm. Although pre- and post-processing MAR may improve CBCT images by minimizing metal artifacts, there are always some missed data that cannot be recreated.18, 19

Based on our findings, with the no-MAR setting, considerable artifacts were generated due to dental implant materials. However, there was no detectable difference in artifact intensity between the 2 MAR settings (pre- and post-acquisition).

It has been stated that the amount of artifact generation in CT/CBCT has a relationship with the atomic number of the element. Sancho-Puchades et al.4 determined that ZrO₂ implants produced the most artifacts in CBCT images, followed by Ti-Zr and Ti implants, and Pekkan et al.20 found that the radiopacity of ZrO₂ was more than twice as high as that of Ti. The radiopacity of a material is based on multiple factors such as atomic number, shape, dimension, density, and some other physical attributes. Of these factors, the atomic number (Z) of an element is the factor with the highest impact on its opacity. An element with a greater atomic number produces more artifacts.4 Thus, the differences in the severity of artifacts in CBCT images of Ti-Zr, Ti, and ZrO₂ can be explained by differences in their atomic numbers: titanium (Z: 22), zirconium (Z: 40), and oxygen (Z: 8).13

Despite the increased use of ceramic material due to its several advantages, such as tissue-friendliness, natural color, high bonding strength, and hardness, Demirturk Kocasarac et al.21 showed that ZrO₂ implants underachieved in comparison with Ti and Ti-Zr implants because they generated more artifacts in CT and CBCT scans. Sancho-Puchades et al.4 also stated that Ti implants generated less artifacts than ZrO₂ implants in CBCT images. The results of this study suggest that the Ti-Zr alloy generated less intense artifacts than the ZrO₂ alloy, while the artifacts generated by Ti were intermediate between the other 2 implant materials, but was not statistically significantly different from either.

Moreover, different kVp settings influence the intensity of artifacts in CBCT scans to a remarkable extent. Esmaeili et al. and Draenert et al. found that higher kVp settings created fewer artifacts. Higher kVp settings generate more energetic and numerous photons that hit the detectors, leading to fewer artifacts. Esmaeili et al.22 compared dental implant-generated artifacts in 2 CBCT machines, while Draenert et al.23 compared the dental implant-associated artifacts produced by CBCT and 4-row multi-detector CT. According to Schulze et al., the artifacts generated by dental implants were reduced by using higher kVp settings. However, Haramati et al.24 obtained no advantage by using a high kVp setting to refine the image quality around the metal prosthesis in CT scans. Nevertheless, our study showed that the higher kVp setting resulted in lower artifact intensity. Various methods have been used to reduce metal artifacts in different CBCT machines; however, further studies are needed to establish whether their application is clinically reliable for eliminating artifacts.

Despite efforts to properly place the implants within the machines and to use standard implant protocols to mimic in vivo conditions, the in vitro nature of this study was still a limitation. In vivo, the voxel value measurements are influenced by nearby anatomic structures, while the geometric distribution and intensity of artifacts for imaging techniques are only affected by the experimental setup in vitro. In actual clinical cases, several factors (e.g., patient movement during the scan, and metallic artifacts due to prostheses) might hinder the diagnosis. Other limitations are the limited sample size of 4 implants and the use of only 1 CBCT unit. Hence, future studies are needed for in vivo testing with different types of machines and a larger sample.

In conclusion, this in vitro study suggests that even though MAR decreased the intensity of image artifacts, no significant difference in terms of artifact generation was found according to whether the algorithm was applied before or after image acquisition. For instance, if the operator does not turn on the algorithm prior to acquisition for some reason, he/she can turn it on after image acquisition and still obtain the same amount of MAR. In addition, artifacts can be minimized by using a Ti-Zr alloy, in comparison to Ti and ZrO₂ alloys, at a high kVp setting.