Difference in the artefacts production and the performance of the metal artefact reduction (MAR) tool between the buccal and lingual cortical plates adjacent to zirconium dental implant

Abstract

Objectives:

To quantify the artefacts production and the performance of the metal artefact reduction (MAR) tool, enabled before or after the acquisition, in cortical plates (buccal and lingual) and other regions adjacent to zirconium implants.

Methods:

Cone beam CT scans were acquired using the OP300 Maxio unit before (control group) and after (implant group) the insertion of a zirconium implant into the posterior region of a dry mandible. Three conditions of MAR tool were tested: “without MAR”, with “MAR activated after acquisition”, and with “MAR activated before acquisition”. The standard deviation (SD), contrast-to-noise ratio (CNR) and voxel values were calculated in the buccal and lingual cortical plates, medullary bone and water region, close to the implant. The structural bone analysis was performed in the medullary bone close to the implant.

Results:

Overall, in control and implant groups, the lingual cortical had higher SD, lower CNR and lower voxel values than the buccal cortical, regardless of the MAR condition (p < 0.05). Implant caused higher SD values and lower voxel values in adjacent regions “without MAR” (p < 0.05). MAR activation decreased SD and changed voxel values when the implant was present, regardless of MAR activation mode (p < 0.05). The activation of MAR increased the trabecular thickness values for the implant group (p < 0.05).

Conclusion:

The expression of artefacts adjacent to zirconium implants is greater in the lingual than in the buccal cortical. The greater the expression of artefacts in this region, the greater the effectiveness of the MAR tool in homogenizing the grey values, regardless of the time of its activation.

Introduction

Cone beam CT (CBCT) images might be affected by different types of artefacts, such as the ones caused by beam hardening, photon starvation, partial volume effect, undersampling, exomass, detector miscalibration, and patient motion. All of these can cause alterations in the resulting images that do not represent the original structure, sometimes being so degrading to image quality that makes it diagnostically unusable.^1,2

Both beam-hardening and photon starvation artefacts are caused mainly by the interaction of the X-ray photons with a high atomic number material, which results in filtering of the incident beam, removing the lower energy photon. This increases the mean beam intensity in case of the beam-hardening phenomenon, or even filters the beam entirely, in case of the photon starvation.^1,3

The influence of these artefacts in diagnostic tasks is well described, such as in mimicking root fracture lines^4,5 or hindering viewing the correct bone levels around metal implants,^6,7 for example. Also, zirconium implants have a higher atomic number than titanium implants, and therefore, the artefacts generated by it can cause an even higher degradation in image quality than its metallic counterpart.^8–10

Metal artefacts reduction (MAR) algorithms have been developed and tested to improve the visualization of structures around these materials.^11–16 Generally, MAR algorithms are effective in technical image quality parameters when evaluated objectively, reducing the standard deviation of the voxel values and increasing the contrast-to-noise ratio (CNR) of these images, but there is controversy when evaluating its effectiveness in different diagnostic tasks.^4,9,15,17,18

Many units present this MAR tool, but most of them have in common that the algorithm has to be enabled before the acquisition occurs. But it is not known what changes the use of this tool does to the acquisition phase, which leads to believe that it is an algorithm applied on the basis images during reconstruction, and not on the way they are acquired. With this, Cranex 3Dx (Soredex Oy, Finland) and OP300 (Instrumentarium Dental, Finland) units can enable this tool before or after the acquisition. The volume can be acquired and reconstructed with the tool turned on from the beginning, or acquired with the tool turned off, and then the image may be retrieved and reconstructed again with the algorithm enabled later. Given the uncertainty of how this tool works, it is important to know if there is any difference between these options so that practitioners and researchers can use them safely according to their needs.

The regions closer to the high-density objects are the most affected by artefacts, however, it is still unclear whether there is a difference in the expression of artefacts in different bone cortical plates. Also, the possible difference between the activation of MAR before or after the acquisition has not been studied. Therefore, the aim of this study was to evaluate the influence of the MAR tool, enabled before or after the acquisition, in cortical plates (buccal and lingual) and other regions around a zirconium implant. The null hypothesis was that there would be no differences in the production of artefacts between the buccal and lingual cortical nor between the use of the MAR or the moment of its activation.

Methods and materials

The local research ethics committee approved this research without any restriction under the protocol number CAAE no 70218017.4.0000.4118.

Sample preparation

One partially edentulous dry human mandible was used as a phantom. The region of the tooth 46 was prepared by an operator with experience in implant dentistry to receive a zirconium oxide dental implant (4 × 11 mm, Z-Look3, Z-systems, Switzerland). Additionally, two epoxy resin-based (ERB) tissue substitute blocks (18 × 10 × 7 mm) were fixed with wax: one on the buccal cortical plate on the right posterior region and another on the anterior region aligned to the middle line. These blocks served as a reference for the standardized selection of axial images that were used for the quantitative analysis of the expression of artefacts produced.

CBCT acquisitions

The phantom was scanned on the OP300 Maxio unit (Instrumentarium Dental, Tuusula, Finland). The energetic parameters, voxel size and field of view (FOV) remained constant in all the acquisitions: 8 mA, 90 kVp, 0,2 mm and 6 × 8 cm. Regarding the use of MAR tool, this unit allows three options in relation to its use: non-activation, activation after scan and activation before the scan. In the present study, the images were initially acquired and reconstructed without the MAR tool activation (“without MAR” condition). Posteriorly, the MAR tool was applied after the scans without the need to obtain new acquisitions (“MAR after” condition). Finally, the images were acquired with the activation of the MAR tool before CBCT acquisition (“MAR before” condition).

The mandible was placed inside a cylindrical plastic container (diameter of 16 cm) and kept fixed with an impression material. This container was filled with water to simulate attenuation of the X-ray beam by the soft tissues. The position of the phantom was standardized by light guides of the unit in the center of the FOV, and it was not touched or moved between all the CBCT scans. The images were acquired before and after implant insertion, composing two main groups of images sets: control group (without implant) and zirconium group (with implant). Figure 1 shows examples of axial views according to groups and MAR tool conditions. Each scan protocol was repeated five times considering the study factors: MAR tool conditions (without MAR, MAR after and MAR before the acquisitions) and experimental groups (control and implant), totalizing 30 CBCT scans (3 MAR conditions × 2 groups × 5 acquisitions).

Figure 1. .

Examples of cropped axial views of the control group and zirconium group, according to MAR conditions available in OP300 Maxio unit. MAR, metal artefact reduction.

Image analysis

Image evaluation was performed by an oral radiologist, under dim light condition. For each volume acquired, it was selected the first axial image in which was possible to visualize the upper level of the ERB block positioned close to the implant. This block is indispensable for standardized selection of the images to be evaluated, especially for a set of images without dental implant.

Cortical plates

The selected axial image was exported for evaluation in the ImageJ software (National Institutes of Health, Bethesda, MD, USA). To establish the regions of interest (ROIs), two lines were determined (Figure 2): the first line parallel to the long axis of the mandibular body on the right side passing through the center of the implant (Figure 2A) and a second line perpendicular to the first one (Figure 2B). Then, two square ROIs of the same size (2 x 2 mm) were drawn: one in the buccal cortical plate and another in the lingual cortical plate of the mandible intercepting the perpendicular line (Figure 2C). The pre-established ROI size was as large as possible to cover the cortical plates, but neighbouring structures were avoided to prevent partial volume effect. The quantitative analysis of the artefact production was performed from the measurement of standard deviation (SD) of the grey values and voxel values. In addition, the CNR was calculated. To measure the CNR values, an additional ROI of the same size as the others was established in the most posterior region of the axial image evaluated (Figure 2C), once it is an area theoretically free of the artefact’s effects. This additional ROI is required to perform this calculation according to the formula described by Bechara et al.¹¹

Figure 2. .

Determination of the regions of interest to evaluate the impact of artefacts production on mandibular cortical plates. (A) A line was drawn parallel to the long axis of the mandibular body on the right side and in the center of the zirconium implant. (B) After this, another line was determined perpendicular to the first line. (C) Three square ROIs of 2 × 2 mm were established in the intersection between the perpendicular line and in the regions close to the zirconium implant: one located on the buccal cortical plate (ROI 1) and one in the lingual cortical plate (ROI 2), and another over the water close to the lingual cortical plate and to the zirconium implant area (ROI 4). In the medullary bone region close to zirconium implant, one more ROI with the same size and shape was established (ROI 5). Two additional control ROIs (ROI 3 and 6) were established for calculating CNR: one in the most posterior region over the water and another in the ERB block located in the anterior mandibular region. CNR, contrast-to-noise ratio; ERB, epoxy resin-based; ROI, region of interest.

$${\displaystyle CNR=\frac{\left|Mea{n}_{implant}-Mea{n}_{Control}\right|}{\sqrt{S{D}_{implant}^2+S{D}_{control}^2}}}$$

Regions close to the implant

In the same axial image selected previously, two other square ROIs (2 × 2 mm) were determined: one over the medullary bone close to the implant area, and another over the water close to the lingual cortical plate and to the implant area (Figure 2C). The same measurements (SD, CNR and voxel values) were obtained. To measure the CNR values, a ROI with the same size was established in the ERB block located in the anterior mandibular region (Figure 2C).

Structural bone

The CBCT scans were exported for evaluation in the ImageJ/Fiji software (National Institutes of Health, Bethesda, MD, USA). In the axial image where the upper level of the ERB block was visualized, a ROI (2 × 2 mm) was determined in the same medullary bone region close to the implant area used in the previous evaluation. Thus, each image was cut to 2 × 2 mm (10 × 10 voxels). In the coronal/sagittal plans, the images were cropped according to the length of the phantom. Based on previous studies,^19,20 all scans were thresholded and converted into a binary image from the “Moments” automatic binarization method. From this, using the BoneJ plugin (http://www.bonej.org), fractal dimension, connectivity density, trabecular thickness (Th. Th.), and trabecular spacing (Tb. Sp.) were measured.

For all analyses, the ROIs positions were standardized using the ROI manager function of the ImageJ and ImageJ/Fiji software. The analyses were performed in 8-bit images.

Statistical analysis

The analysis was performed using SPSS v. 24.0 (IBM Corp., Armonk, NY) and GraphPad Prism v. 7.0 (GraphPad Software, La Jolla, CA) software, with a significant p-value < 0.05. The power analysis was 0.9 based on the mean minimum differences among the groups studied, the mean standard deviation and the number of scans per group calculated in Biostat software (v. 5.3, Instituto de Desenvolvimento Sustentável Mamirauá, Tefé, MA, Brazil).

The SD, CNR, and voxel values were compared by analysis of variance (multiway ANOVA) with post-hoc Tukey test, in order to test the main effects of the factors studied (presence of implant and MAR condition). In addition, for cortical plates analysis, another factor studied was the region (buccal or cortical). The ANOVA was conducted independently for ROIs located on the medullary bone and over the water close to the implant area. Regarding the structural analyses, ANOVA compared the results to test the main effects of the factors studied (presence of implant and MAR condition).

Results

The mean of SD, CNR and voxel values on buccal and lingual cortical plates are shown in Table 1. The buccal ROI differed from lingual ROI, regardless of the presence of the dental implant (p < 0.05). Overall, the lingual ROI showed higher SD and lowers values of CNR and voxel. In relation to MAR condition, for the buccal ROIs of the control and implant group, SD values did not differ significantly in all MAR condition (p > 0.05). The same occurred for the lingual ROI of the control group (p > 0.05). On the other hand, the SD values of lingual ROI in the implant group decreased significantly with the MAR activation (p < 0.05), regardless of the MAR activation mode. For CNR values, there were no significant differences between MAR condition in both groups (p > 0.05). In opposite, the voxel values of both cortical plates differ significantly between the acquisition without MAR and with MAR activated for the implant group (p < 0.05). In the presence of the implant, the activation of MAR induced a slight whitening in the buccal cortical. However, the opposite occurred in the lingual cortical. When comparing control and implant groups in each region, it was observed that the presence of the implant significantly increased the SD values only in the lingual cortical plate, regardless MAR condition (p < 0.05). On the other hand, the presence of the implant decreased significantly the CNR and voxel values in both cortical plates, regardless of the MAR condition (p < 0.05).

Table 1. .

Mean of SD, CNR and voxel values of the control and implant groups according to the ROIs located in the buccal or lingual cortical plates and MAR tool activation conditions

	Groups	ROIa	MAR condition
			Without MAR	MAR after	MAR before
SD	Control	Buccal	7.25 (0.93)	7.25 (0.93)	7.64 (0.67)
		Lingual	12.23 (0.61) b	12.23 (0.61) b	11.96 (0.63) b
	Implant	Buccal	7.33 (1.86)	7.07 (0.95)	6.58 (1.12)
		Lingual	14.68 (2.63) A	10.49 (1.40) B	9.93 (1.29) B
CNR	Control	Buccal	12.25 (1.71) b	12.25 (1.71) b	11.79 (0.92) b
		Lingual	6.08 (0.40) b	6.08 (0.40) b	6.26 (0.39) b
	Implant	Buccal	6.15 (1.36)	6.90 (0.82)	7.60 (1.27)
		Lingual	3.32 (0.54)	3.68 (0.39)	3.73 (0.33)
Voxel value	Control	Buccal	146.21 (2.34) b	146.21 (2.34) b	145.95 (0.67) b
		Lingual	129.05 (2.19) b	129.05 (2.19) b	128.89 (1.56) b
	Implant	Buccal	97.98 (2.12) A	102.91 (1.28) B	105.03 (2.59) B
		Lingual	100.60 (3.44) A	91.20 (0.73) B	90.31 (1.12) B

CNR, contrast-to-noise ratio; MAR, metal artefact reduction; ROI, region of interest; SD, standard deviation.

Different letters indicate statistical difference between the acquisitions without MAR, with MAR before and with MAR after.

^aBuccal ROI differed from lingual ROI regardless the presence of dental implant.

^bDiffered significantly from “implant group” within each ROI.

The mean of SD, CNR, and voxel values of the control and implant groups in the ROIs located in the water adjacent to the lingual cortical plate and in the medullary bone adjacent to implant area are shown in Table 2. For both ROIs, the implant group showed significantly higher SD values in the acquisitions without MAR activation (p < 0.05). Regarding the CNR values, they decreased significantly when the implant was present in ROI adjacent to the implant area when MAR was not activated or activated before CBCT acquisition (p < 0.05). Voxel values of implant group differed significantly from the control group for all MAR condition in both ROIs (p < 0.05), except adjacent to the lingual cortical plate with MAR activated after CBCT acquisition. The effect varied since the presence of implant made the area slightly darker adjacent to the lingual cortical plate and slightly brighter in medullary bone, except for the group without MAR. For the control group, there were no significant differences between the MAR condition (p > 0.05). In opposite, for the implant group in the both ROIs, the activation of MAR decreased significantly the SD values and increased the voxel values, regardless of the MAR activation mode (p < 0.05). The exception for this was observed in the adjacent lingual cortical plate ROI, since that occurred a slight reduction of the voxel values between MAR activated after and MAR activated before the CBCT acquisitions of the implant group. Regarding CNR values, there was no effect of MAR in both control and implant groups (p > 0.05).

Table 2. .

Mean of SD, CNR and voxel values of the control and implant groups in ROIs located in the water adjacent to lingual cortical plate and in medullary bone adjacent to implant area, according MAR tool activation conditions

	ROIa	Group	MAR condition
			Without MAR	MAR after	MAR before
SD	Adjacent to lingual cortical plate	Control	3.07 (0.39) b	3.07 (0.39)	3.32 (0.52)
		Implant	4.64 (1.32) A	3.24 (0.82) B	3.64 (0.48) B
	Medullary bone	Control	6.27 (0.98) b	6.27 (0.98)	6.10 (0.73)
		Implant	11.41 (1.29) A	5.78 (0.66) B	6.86 (0.76) B
CNR	Adjacent to lingual cortical plate	Control	13.89 (1.51)	13.89 (1.51)	12.54 (0.43)
		Implant	12.61 (2.85)	14.61 (2.46)	12.94 (1.25)
	Medullary bone	Control	7.92 (0.96) b	7.92 (0.96)	7.79 (0.59) b
		Implant	5.90 (0.59)	7.01 (0.79)	5.98 (0.55)
Voxel value	Adjacent to lingual cortical plate	Control	48.90 (0.75)b	48.90 (0.75)	48.91 (0.92) b
		Implant	45.18 (1.21) C	48.44 (0.52) A	46.91 (0.61) B
	Medullary bone	Control	59.58 (0.67) b	59.58 (0.67) b	59.66 (0.55) b
		Implant	47.19 (1.98) B	70.00 (0.25) A	70.33 (0.65) A

CNR, contrast-to-noise ratio; MAR, metal artefact reduction; ROI, region of interest; SD, standard deviation.

Different letters indicate statistical difference between the acquisitions without MAR, with MAR before and with MAR after.

^aROIs were not compared to each other.

^bDiffered significantly from “implant group” within each ROI.

The structural analyzes are shown in Table 3. The fractal dimension and Tb. Sp. values were not influenced by the factors studied (implant and MAR) (p > 0.05). For connectivity density values, the implant group showed significantly lower values when compared to the control group, regardless of the MAR condition (p < 0.05). For Tb. Th. values, the control group did not differ to the implant group without the MAR activation (p > 0.05). In opposite, there was a significantly increased in Tb. Th. values when the implant was present in both MAR activation mode (p < 0.05). For the implant group, the activation of MAR increased significantly the Tb. Th. values (p < 0.05).

Table 3. .

Structural analyses in the ROI located in medullar bone adjacent to implant area, according MAR tool activation conditions

Analysis	Group	MAR condition
		Without MAR	MAR after	MAR before
Fractal	Control	2.56 (0.00)	2.56 (0.00)	2.56 (0.00)
	Implant	2.49 (0.00)	2.49 (0.00)	2.49 (0.00)
Connectivity density (mm³)	Control	0.016 (0.00) a	0.016 (0.00) a	0.017 (0.00) a
	Implant	0.009 (0.00)	0.008 (0.00)	0.008 (0.00)
Tb.Th mean (mm)	Control	5.77 (0.24)	5.77 (0.24) a	5.68 (0.08) a
	Implant	5.94 (0.09) C	7.09 (0.35) A	6.67 (0.36) B
Tb.Sp mean (mm)	Control	4.03 (0.16)	4.03 (0.16)	4.06 (0.12)
	Implant	4.23 (0.11)	3.84 (0.42)	4.08 (0.17)

MAR, metal artefact reduction; ROI, region of interest.

Different letters indicate statistical difference between the acquisitions without MAR, with MAR before and with MAR after.

^aDiffered significantly from “implant group” within each ROI.

Discussion

The present study showed that the expression of artefacts adjacent to a dental implant located in the posterior region of the mandible is greater in the lingual cortical than in the buccal cortical. To our knowledge, previous studies have evaluated the production of artefacts in the cancellous and/or cortical bone,^8,21 but not by comparing the two cortical plates. As artefacts related to dental implants can impair measuring the cortical bone thickness⁷ and the assessment of peri-implant bone defects,^14,16 our result is interesting because it presents the insight that there may be differences in the interpretation of these clinical conditions depending on the cortical plate evaluated.

The different expression of the artefacts between the cortical plates may be related to the difference in anatomical positioning of the buccal and lingual cortical in relation to the X-ray beam and the dental implant. When X-rays are laterally to the patient, structures closer to the outer surface of the skull such as buccal cortical plate are passed by a high signal-to-noise ratio, because of little other tissues are in the X-ray path. Conversely, when X-rays are posterior to the patient, the X-rays pass the lingual cortical plate first; however, those X-rays have already been attenuated by several other tissues such as the tongue and spine. Therefore, in general the lingual bone is more subject to beam hardening (i.e. X-rays passing through it have a higher average energy compared to the buccal bone). All these factors plus the fact that the whole image is subject to exomass effects and possible interferences of the image reconstruction process probably have repercussions in the image quality and in the different artefacts formation between the cortical ones.

Our main objective was not to evaluate the difference among the cortical plates in the absence of the implant (control group); however, it was important to analyze the bone characteristics of the mandible studied, since possible differences in the cortical grey values ​​could exist independently of the presence of the dental implant. From this analysis, it was possible to observe that the presence of the implant significantly increased the SD values ​​of the lingual cortical, and significantly decreased the CNR and voxel values ​​in both cortical plates. In addition, the control group was important because it showed that, in the absence of high-density materials, there was no difference in the SD, CNR, and voxel values results ​​between groups without MAR and with MAR activated, regardless the moment of its activation. In this way, we have confirmed that the MAR tool only acts in the presence of artefact-generator materials as some previous study had found.^9,22 It is also interesting to note that SD, CNR, and voxel values were exactly the same in the conditions “without MAR” and “with MAR after”, while numerical differences were observed between “MAR before” and those groups. The fact that the first two groups were derived from the same CBCT scans, while the “MAR before” group resulted from apart scans may explain those results. Therefore, this difference is not caused by MAR, but is related to the inter sweep fluctuations of tube output, since the images in the conditions “without MAR”/ “MAR after” and “MAR before” were obtained at different times.

In an in vitro model using silicone mandibles, Benic et al²³ evaluated the expression of artefacts in different positions and distances adjacent to titanium implants placed in anterior and posterior regions of the dental arch. These authors found no difference between the buccal and lingual aspects of the implant site, regardless of the implant region.²³ Sancho-Puchades et al,¹⁰ using a similar methodology, tested different types of dental implants and observed that zirconium implants generate significantly more artefacts than titanium and titanium–zirconium implants, also without differences between the buccal and lingual aspects of the implant site. Such contrasting findings can be attributed to the size and location of the studied regions (one pixel and 0.5 to 2 mm apart) and to the use of silicone models rather than a human mandible, as done in the present study.

The increase in energetic parameters and/or spatial resolution of the image have been pointed out as effective alternatives to improve the performance of the CBCT in the diagnosis of several conditions, as in the detection of peri-implant dehiscences and fenestrations⁶ and in the detection of root fractures.^5,24 On the other hand, although the MAR tool improves image quality,^9,11 its activation has demonstrated controversial results in these same diagnostic tasks.^4,14,15 We observed that the MAR significantly reduced SD values ​​only in the lingual cortical, where the artefacts were more pronounced. Thus, we suggest further clinic studies comparing the impact of the use of MAR on cortical plates bone alterations, in order to be able to indicate the activation or not of this tool in CBCT exams carried out to evaluate bone tissue adjacent to dental implants.

In general, corroborating with the results obtained in the cortical plates evaluation, the presence of the dental implant reduced the image quality in other regions close to the implant area, since it significantly increased the SD values, decreased the CNR and voxel values. In addition, when the artefacts were more expressive, the MAR tool acted causing a significant decrease in the variability of the grey values. Regarding structural bone analysis, previous studies evaluated the influence of several factors on these measures, such as exposure parameters,²⁰ scan parameters,²⁵ voxel size²⁰ and position of the object in the different FOVs.²⁶ However, for the first time, the present study evaluated the influence of the artefacts generated and the MAR tool in that analysis. In general, the presence of the implant significantly decreased the connectivity density ​​and increased the Tb. Th. values. Regarding the MAR tool, its activation significantly increased only the Tb. Th. values. As a pioneer study, it is not possible to compare our results to others, but they can serve as a reference for future researches. Professionals and researchers that use structural analyses should be aware that artefacts and MAR can influence some aspects of the evaluation.

The possibility of obtaining CBCT exams with and without MAR (two volumes reconstructed from a single acquisition) is a new feature present in some recently released CBCT units. As far as we know, our study is the first to investigate if the moment of activation of the MAR tool (before or after image acquisition) influences the ​SD and CNR of the grey values. In both the control and implant groups, the moment of MAR activation did not influence the results. Although different CBCT units may present different algorithms for image reconstruction and processing, we believe that this result opens up possibilities for in vivo studies that compare the influence of this tool on diagnostic decisions.

Similar to previous studies,^8,22,27 image evaluation was performed by only one oral radiologist. We believe this is appropriate because the evaluation is done objectively in the ImageJ software by determining the ROIs using the MACRO function. Considering our results, we suggest further studies in different regions of the dental arch and using other high-density materials that trigger the beam-hardening phenomenon, in order to verify the performance of the MAR tool and the pattern of artefact production.

In conclusion, the expression of artefacts adjacent to zirconium implants is greater in the lingual cortical than in the buccal cortical. Also, the greater the expression of artefacts in this region, the greater the effectiveness of the MAR tool in homogenizing the grey values, whether it is activated before or after the acquisition of CBCT images.