The detection of vertical root fractures in post-core restored teeth with cone-beam CT: in vivo and ex vivo

Abstract

Objectives:

To compare the accuracy of cone-beam CT ex vivo and in vivo for the detection of artificially created large and small vertical root fractures in extracted teeth restored with post-core.

Methods:

Individual metal cast post-cores were fixed in the root canals of 50 extracted single-rooted human teeth. In 30 teeth fractures were created by tapping posts with a hammer. The teeth were sterilised in autoclave and embedded into bite-plates made of silicon impression material. Cone-beam CT scanning was performed ex vivo and in vivo . For the in vivo scanning, teeth in sterile plastic bags were inserted into the mouths of volunteers. Then the teeth were sectioned with low-speed saw and the widths of the VRFs were measured microscopically. The teeth were distributed into 2 groups in accordance with the measured fractures’ widths: large (wider than 180–250 µm) and small (80–150 µm). Five observers assessed the presence of vertical root fractures on axial CBCT slices. Sensitivity, specificity, accuracy and inter examiner agreement were calculated.

Results:

The accuracy of cone-beam CT in vitro for large and small vertical root fractures detection was 0.56 and 0.40 respectively (p = 0.043). The sensitivity values were 0.53 and 0.27 for large and small vertical root fractures, respectively (p = 0.043). The visualisation of fracture lines in vivo was impossible in 90 % of cases, because of low image quality. Inter examiner reliability analysis showed κ values ranging from 0.02 to 0.54.

Conclusions:

Fracture width affected the in vitro detectability of vertical root fractures by cone-beam CT in teeth with metal cast post-cores. The detectability of root fractures in vivo was decreased because of low image quality, making the assessment of sound tooth tissue impossible.

Introduction

American Association of Endodontists defined “vertical root fracture” (VRF) as a longitudinally oriented fracture of the root that originates from the apex and propagates to the coronal part.¹ An overall prevalence of VRFs from 3 to 5% has been reported,^2–5 with the majority of fractures observed in endodontically treated teeth. The aetiology of VRFs is not clearly understood and is considered to be multifactorial. Factors associated with endodontic treatment are recognised, such as changes in mechanical properties of root dentin due to the removal of the pulp, loss of sound tooth structures and use of posts-retained constructions to restore the teeth.⁶

Cone-beam CT (CBCT) is proposed as a method which could aid in the diagnosis of VRFs.⁷ However, metal constructions including intracanal metal posts, which are commonly used for the restoration of endodontically treated teeth, cause artefacts⁸ which can obscure fracture line. According to the previous in vitro studies radiopaque materials decreased the diagnostic value of CBCT for VRFs detection.^9–17 At the same time, it has also been shown that the accuracy of VRFs detection depend on the width of the fracture^18,19 and is lower in vivo compared to in vitro.¹⁸ A living patient undergoes slight unconscious movements (breathing, heartbeats) which affect the image quality by decreasing the actual spatial resolution of the acquired image.²⁰

The aim of the present study was to compare the performance of cone-beam CT in vitro and in vivo for the detection of artificially created large and small vertical root fractures in extracted teeth restored with post-core.

methods AND Materials

50 intact single-rooted human teeth extracted due to periodontal or orthodontic indications were chosen for the study. The root canals of teeth were prepared with the use of manual K-files (Dentsply/Maillefer, Ballaigues, Switzerland) and post spaces were prepared with gates glidden drills (Dentsply/Maillefer, Ballaigues, Switzerland) (Figure 1). The teeth were immersed in a benzalkonium chloride disinfectant solution (“Diabac”, INTERSAN-plus, Mitishchi, Russia) for 45 minutes and autoclaved (Prestige Medical Limited, Coventry, UK) at 132°C for 20 minutes.

Figure 1.

Experiment design (adapted from Makeeva et al. 2016): after creating of a fracture by a hammer, 1 – the tooth is sterilised, 2 – embedded in a plastic cement and post-core inserted, 3 – embedded in impression material to create a bite-plate; 4,5 – Cone-bean CT scanning is performed in vitro and in vivo respectively; 6 – root slice is prepared; 7 – microscopic evaluation of a fracture width is performed.

Individual metal cast post-cores were manufactured out of cobalt-chrome alloy. In 30 teeth VRFs were created by tapping posts with a hammer. Then all teeth were inspected with the dental operating microscope OPMI PROergo (Carl Zeiss Meditec AG, Jena, Germany); the presence of VRFs in the fracture group and the absence of fractures in non-fractured group was confirmed. The teeth were placed in plastic cylinders cut from disposable syringes and filled with resin (Rebaron, GC Corporation, Tokyo, Japan). After setting of the resin, plastic cylinders were fully imbedded in the silicon impression material and the bite-plates were formed. CBCT scanning of each sample ex vivo was performed.

The approvals from the institutional review board and ethical committee were obtained (protocol №09-14). 30 patients who were to undergo CBCT scanning participated in the study. The written informed consent was obtained from all patients. The teeth in bite plates were rinsed with anticeptic, placed in clear plastic contamination barrier bags and inserted in the mouths of the patients beside the anterior teeth (as previously described by Makeeva et al. 2016)¹⁸; then, CBCT scanning was performed in vivo.

Samples ex vivo were scanned using the 3D Accuitomo 170 (3D Accuitomo; J. Morita Mfg. Corp., Kyoto, Japan) with the following parameters: FOV of 8х8 cm3, voxel size (0.16 mm)3, 80 kV, 4 mA, 30.8 s. The teeth were scanned individually in the acrylic resin and bite-plates, in vertical position; no phantom and scatter equivalent was used. Patients were scanned using a 3D Accuitomo 170 unit (3D Accuitomo; J. Morita Mfg. Corp., Kyoto, Japan), FOV 8х8 cm3, voxel size (0.16 mm)³, 90 kV, 4–5 mA, 30.8 s.

After scanning, the widths of the fractures were measured. The teeth were removed from the bite-plates and horizontaly sectioned in the middle third of the root with a slow-speed saw (Isomet, Buehler, Lake Bluff, IL). The images of each root section were acquired with a DCM-800 camera (8х magnification) attached to a Neophot 2 stereomicroscope. ArcSoft PhotoStudio 5 (ArcSoft Inc., Fremont, CA) was used for the processing of the acquired images. The widths of VRFs were measured as the largest distances between the edges of the fractures. Evaluation of root sections provided the opportunity to determine the exact width of the VRF on the certain level of each sample, not only on the root surface. 20 roots without VRFs were also sectioned and photographed in order to confirm the absence of the fractures. The widths of the created VRFs varied from 80 to 250 µm. Of 30 root sections, in 13 VRFs were incomplete (80–150 µm wide) and in 17 VRFs were complete (≤150 µm).

For each tooth 5 axial CBCT slices from ex vivo and in vivo scanning were chosen, which corresponded to the level of the horizontal sections used for fracture width measurement. These axial slices were distributed into 2 groups according to the fracture widths. For non-fractured teeth, random 5-slices series from the middle third of the root were chosen. The resulting axial CBCT slices were exported in jpg format and randomly distributed in a PowerPoint presentation. The presentation was displayed on a monitor of a laptop (Apple MacBook Pro 15’’, Apple Inc., Cupertino, CA). Five blinded calibrated observers evaluated the presence of VRFs on the slice series during individual sessions with the use of 3-grade scale (definitely a fracture; definitely not a fracture; uncertain). To calibrate the observers, 10 images with metal post-cores (5 with fracture and 5 without) and 10 images without post-cores (5 with fracture and 5 without), other than used in the study, were shown. Particular attention was drawn to the differences between artefacts from metal constructions and fracture lines. Besides, all observer had previous experience in diagnosing VRFs for at least 5 years.

For the statistical analysis the data were imported in SPSS (IBM SPSS Statistics Version 22, Armonk, NY). The mean and median values of specificity, sensitivity and accuracy of CBCT for VRFs detection were calculated in each group. Accuracy was calculated as the proportion of true positive results (true positive and true negative) in the study sample. Wilcoxon signed-rank test was used for the assessment of statistical significance of differences between groups. Kappa-coefficient was used to evaluate inter examiner agreement.

Results

CBCT accuracy values for the detection of large and small VRFs in vitro were 0.56 and 0.40 respectively (p = 0.043) (Table 1). The sensitivity values of CBCT were 0.53 and 0.27 for large and small VRFs respectively (p = 0.043). No significant differences in CBCT specificity were found between VRF widths (p = 0.345). For the in vitro CBCT specificity and accuracy, the power analysis showed 99 and 98% power, respectively, for the groups with different fracture widths (Type 1 error rate 5%). The lack of statistical significance between specificity values is quite logical as the specificity (true negative rate) measures the proportion of actual negatives that are correctly identified as such (e.g. the percentage of teeth without fractures which are correctly identified as not being fractured). That is, specificity cannot depend on fracture widths, as it is calculated from the non-fractured teeth.

Table 1.

Sensitivity, specificity and accuracy of cone-beam CT (CBCT) scanning in vitro: comparison between control teeth with no defects and teeth with fractures of different widths

Observer	50–150 µm			>150 µm
	Sensitivity	Specificity	Accuracy	Sensitivity	Specificity	Accuracy
1	0.22	0.22	0.22	0.64	0.70	0.67
2	0.22	0.56	0.39	0.57	0.30	0.46
3	0.11	0.89	0.50	0.43	0.80	0.58
4	0.44	0.56	0.50	0.64	0.78	0.70
5	0.33	0.44	0.39	0.36	0.50	0.42
Median	0.22	0.53	0.39	0.57	0.70	0.58
Mean	0.27	0.56	0.40	0.53	0.62	0.56
p-value				0.043a	0.345	0.043a

^aDifferences between groups statistically significant (p < 0.05).

It was impossible to assess the presence of VRFs in 90% of cases in vivo (all observers graded the images as “uncertain”) because of strong artefacts caused by metal posts and low image quality (Figure 2E and L).

Figure 2.

Comparison of teeth with 100 µm (A-G) and 200 µm (H-N) fractures’ widths: photographs (A, B, H, I), the axial cone-beam CT slices at the same level without metal post in vitro (C, J), with metal post in vitro (D, K); with metal post, in vivo (E, L); digitised images of the coronal surfaces of root slices at the same level at 2x magnification (F, M) and 8x magnification (G, N).

Inter examiner reliability analysis showed κ values ranging from 0.02 to 0.54 which corresponded to “slight” to “fair” agreement for different pairs of observers (Table 2).

Table 2.

Cohen’s κ coefficient for the different pairs of observers

Pair of observers	Cohen’s κ
Observers 1 and 2	0.25
Observers 1 and 3	0.53
Observers 1 and 4	0.54
Observers 1 and 5	0.02
Observers 2 and 3	0.34
Observers 2 and 4	0.44
Observers 2 and 5	0.02
Observers 3 and 4	0.44
Observers 3 and 5	0.14
Observers 4 and 5	0.02

Discussion

In the present study, the accuracy of CBCT for the detection of VRFs of different widths was assessed in vivo and ex vivo in teeth restored with metal cast post-cores. It was the first study to assess VRF detection accuracy in teeth with metal cast post-cores in which the width of the fractures was measured. We used the method described by Makeeva et al. (2016), which allowed to compare the same fractures ex vivo and in vivo without the changes of fracture widths, which inevitably occur if the tooth was first imaged and then extracted. We chose CBCT slices, which corresponded accurately to the root sections that were used for fracture width measurements. Although the assessment of only 5 axial slices is a clinically unrealistic scenario, it allowed to know the exact width of the fracture corresponding to the slices on which it was detected. The division of fractures into “small” and “large” was performed depending on the propagation of the fracture: complete (large) fractures were 180–250 µm wide while the incomplete (small) fractures were 80–150 µm wide. In the study by Patel et al. (2013) incomplete (50–110 µm) and complete (more than 200 µm) fractures were also compared.¹⁹ The authors used optical coherence tomography to assess the fracture widths on the tooth surface. The differences in fracture widths with the study by Patel et al. (2013) might be due to different methods of measurement and sample preparation.

There were significant differences in the values of sensitivity and accuracy between the groups with different fracture widths in vitro in the present study. However, the visualisation of fracture lines in vivo was impossible in 90% of cases, because artefacts obscured the root surface from inspection (Figure 2E and L). As a result, observers were unable to identify the absence or presence of the fracture in the majority of cases. Therefore, no comparison between the values in vivo and ex vivo could be performed.

The study by Hassan et al. (2009) was one of the first to show the influence of radiopaque root filling materials on the specificity of CBCT for the detection of VRFs in vitro.¹¹ Since then a number of studies were published aiming at the comparison of CBCT diagnostic efficacy for the detection of VRFs in teeth restored with metal posts (Table 3).

Table 3.

The results of the previous studies on the detection of vertical root fractures (VRFs) in teeth with metal posts using cone-beam CT (CBCT) in vitro

Author	voxel size (mm)	Type of post	Sensitivity	Specificity	Accuracy	Method of fracture creation
Melo 2013	0.2	metallic	0.41–0.44a	0.87–0.92	0.65–0.68	hammer
Jakobson 2014	0.2	metallic	0.94 (CBCT1); 0.96 (CBCT2)	1.0; 0.94	-	0.20 mm thick discs
Menezes 2016	0.1	metallic	0.62–0.75b	0.25–0.37	0.50–0.56	UTM
Mohammadpour 2014	0.15	titanium	0.91	0.77	0.82	UTM
	0.15	stainless steel	0.82	0.68	0.73
Moudi 2014	0.3	gold-plated	0.81	1.0	-	hammer
Neves 2014	0.08	cobalt-chromium	-	-	AUC:0.57–0.59 (complete); 0.43–0.50 (incomplete)	UTM
Pinto 2017	0.1	gold-alloy	0.53–0.61c	0.78–0.87	0.71–0.76; AUC: 0.78–0.87	hammer
Abdinian 2016	0.2	screw-type	0.7	0.65	0.67	hammer
Moudi 2015	0.3	gold-plated	0.95–1d	0.89–1	-	hammer
The present study	0.16	metal cast post-core	0.27 (small VRFs)	0.56	0.40	hammer
			0.53 (large VRFs)	0.62	0.56	hammer

AUC, area under (receiver-operator characteristic) curve; CBCT, cone-beam CT; FOV, field of view; UTM, universal testing machine; VRF, vertical root fracture.

^aMaximum and minimum values for various dental software programs

^bMaximum and minimum values for the 3 observers

^cMaximum and minimum values for different regimens (74 kV/12 mA; 74 kV/ 10 mA; 74 kV/8 mA; 74 kV/6.3 mA; 70 kV/12 mA; 70 kV/10 mA; 70 kV/8 mA; 70 kV/6.3 mA)

^dMaximum and minimum values for various FOV (16 × 8 cm and 6 × 6 cm)

In the study by Melo et al. (2013) the presence of a gold post significantly reduced the sensitivity and accuracy values of CBCT for the detection of VRFs from 65–73% to 41–43% and from 76–79% to 65–66%, respectively.⁸ In the study by Jacobson et al. (2014) the presence of prefabricated metal post caused the decrease in the specificity of one of CBCT units (NewTom^® 3G).⁹ Moudi et al. (2014) reported that prefabricated posts also reduced the accuracy of the CBCT scans.¹² Mohammadpour et al. (2014) have shown that the sensitivity, specificity, and accuracy values for VRF diagnosis were significantly lower in teeth with stainless steel and titanium posts compared with those without posts. Interobserver agreement was the highest in the group without posts, followed by stainless steel posts, and titanium posts.¹³

The presence of posts and gutta-percha reduced the sensitivity and accuracy values for the detection of VRFs in the study by Menezes et al. (2016).¹⁴ Neves et al. (2014) found significantly smaller diagnostic value (Area under the ROC Curve (Az)) of CBCT for the detection of VRFs in the group with metal posts, compared with other groups (with fibreglass posts, with gutta-percha and without any filling).¹⁵ Abdinian et al. (2016) reported the reduction of sensitivity, specificity, and accuracy values of CBCT in the presence of gutta-percha or post in the root canal.¹⁶ Pinto et al. (2017) also found that the accuracy and sensitivity of the metallic post group were significantly lower compared with unrestored and fibreglass post groups for the majority of the scanning regimens tested.¹⁷

The reduction of CBCT value for VRF detection in teeth with posts could be due to the artefacts caused by metal constructions. Root filling materials can induce beam hardening effect and therefore produce streaking artefacts, which may either conceal or mimic fracture lines, and therefore, mislead the clinician. This can explain poor inter examiner agreement reported in the present study and in the study by Abdinian et al. (2016).¹⁶ It should be also noted that in the studies cited above, prefabricated or metal cast posts were used, while we used metal cast-post cores which were larger because of the presence of core and therefore could cause stronger artefacts.

The mean specificity, sensitivity and accuracy values obtained for the group of small fractures in the present study ex vivo were generally lower compared with the values reported in the aforementioned studies (Table 3). This may be due to differences in fracture widths. In the previous studies, which assessed VRF detection in teeth with metal posts, the fracture widths were not measured. However, it can be assumed that the fracture width depends on the method used for its creation. The fractures created with universal testing machines under controlled conditions possibly have smaller widths compared with the fractures created with a hammer, while the fractures cut with 0.2 mm thick discs have minimum width of 0.2 mm.

Moudi et al. (2015) reported the decrease in specificity values of CBCT in the presence of a pin in the large-volume group, but not in the small-volume group.²¹ The design of the present study required the use of CBCT with a large field of view, because simultaneous visualisation of the area of interest for the patient and of the experimental sample was accomplished. The sample was placed in the mouth as far from the area of interest as possible, not to impair the quality of its visualisation. Therefore, a larger FOV was used both in vitro and in vivo to obtain comparable results. This is a limitation of the present study, as the use of a smaller voxel size and FOV would have been more reasonable for the detection of VRFs.

The ability of CBCT to visualise small structures or defects (fracture lines) is limited by the quality of the acquired image, which is in turn influenced by spatial resolution, contrast, noise, artefacts and other factors. These parameters are interdependent and therefore should always be considered together.²² Spatial resolution refers to the ability to discriminate small structures in an image. According to Niquist theorem, to guarantee the visualisation with CBCT, a feature size should exceed at least twice the voxel size. For example, if minimum fracture width is 0.1 mm voxel size must be 0.05 mm³ or less. However, ideal situation when a line-pair would be visualised by two pixels (one dark and one bright) is not possible, as there are many factors that further decrease spatial resolution.²⁰ First of all, the edge of the scanned object (i.e. the tooth) is not represented by a single voxel, but is distributed over several adjacent voxels (as different shades of grey). The contrast between the object and background also impacts the possibility of visualisation.²⁰ In case of root fracture the contrast between the fracture space and the tooth seems to be sufficient to favour fracture detection. However, if a voxel is at the junction of two objects of different densities (e.g. tooth and air or soft tissues) it reflects an average value between the true values for these objects (partial volume effect) decreasing the contrast.^23,24 Noise (random variability in voxel values in an image) influences spatial resolution and can also obscure or simulate small details on the image.²² With the improvement of spatial resolution there is an increase of noise and therefore, there is a trade-off between spatial resolution and noise for each reconstruction algorithm. Noise-reduction filters have been developed to reduce image noise while maintaining high-contrast resolution, tending, however, to change the noise texture and sacrifice the low-contrast detectability in the image.²⁵ The image artefacts decrease spatial resolution, and may mislead the clinician by either concealing or mimicking fracture lines. Among the major sources of artefacts in CBCT are X-ray scatter, beam hardening, metal constructions, patient motion, hardware and software limitations and others.²⁶ Metal artefacts can be reduced during reconstruction through the application of artefact reduction techniques.²⁷ The influence of these algorithms on the detection of root fractures was assessed in several studies, however, the results were controversial.^28–30

The analysis of the articles in the review by Brülmann & Shultze (2015) indicated a theoretically maximal spatial resolution of CBCT in vitro of <3 lp mm⁻¹ (median - 2 lp mm⁻¹).²⁰ The widths of the fractures in the present study ranged from 80 to 250 µm, and these fractures were detected in some cases in vitro (mean sensitivity value was 0.27 for small and 0.53 for the large fractures, p = 0.043). However, spatial resolution of the CBCT system in vitro, or “nominal spatial resolution” is different from the actual (reduced) spatial resolution available in living patients.²⁰ For the in vivo scanning, additional image degrading factors play role: greater volume of hard and soft tissues for the X-ray to pass and patient’s motion. It has been shown that movement greater than 0,5 mm is enough to decrease the image quality, while breathing or swallowing can induce movement with amplitude of approximately 6 mm.³¹ Even heartbeat alone induces a slight movement with amplitude of 80 μm.³² These small movements cause motion blur and decrease “nominal spatial resolution”. A realistic detail size of 500 µm has been reported, that could be visualised in a high-quality CBCT of a “visually steady” (i.e. only moving due to heart beat) patient.³³ As a result, it was impossible in the present study to detect 90% of fractures in vivo: the observers stated that they were not able even to assess the presence of fractures due to low image quality.

Based on the results of the present study, vertical fracture lines in teeth restored with metal cast post-cores cannot be accurately identified with CBCT. However, in clinical practice, a diagnosis of VRF would be based on a combination of clinical and radiographic findings.^34,35 The evaluation of the pattern of the bone resorption on CBCT can possibly aid in differentiation between VRF and mimicking conditions.

Conclusion

The detectability of VRFs in teeth with metal cast post-cores by CBCT in vitro depended on its width. The detectability of VRFs in teeth with metal cast post-cores by CBCT in vivo was decreased because of low image quality, making the assessment of sound tooth tissues impossible.