Accuracy in the diagnosis of vertical root fractures, external root resorptions, and root perforations using cone-beam computed tomography with different voxel sizes of acquisition.

AIM: The aim of this study is to assess the accuracy of images acquired with cone-beam computed tomography (CBCT) in the identification of three different root alterations.
MATERIALS AND METHODS: Forty human premolars were allocated to four experimental groups (n = 10): sound teeth (control), vertical root fracture (VRF), external root resorption (ERR), and root perforation (RP). After the root alterations had been produced, four teeth were randomly assembled into 10 macerated mandibles and submitted to CBCT. Images were acquired with five voxel sizes (0.125, 0.200, 0.250, 0.300, and 0.400 mm) and assessed by three experienced dental radiologists. Sensitivity, specificity, positive and negative predictive values, and the areas under the receiver operating characteristic curve (accuracy) were calculated. The accuracy of imaging in different voxel sizes was compared with Tukey exact binomial test (α=5%).
RESULTS: Accuracy with voxel sizes 0.125, 0.200, and 0.250 mm was significantly higher in the detection of ERRs and VRFs than voxel sizes 0.300 and 0.400 mm. No statistical difference was found in terms of accuracy among any of the studied voxel sizes in the identification of RPs.
CONCLUSIONS: Voxel size 0.125 mm produced images with the best resolution without increasing radiation levels to the patient when compared to voxel sizes 0.200 and 0.250 mm. Voxel sizes 0.300 and 0.400 mm should be avoided in the identification of root alterations.

INTRODUCTION

The use of dental radiology and imaging has become essential along all the steps involved in the treatment of dental root alterations, assisting in the diagnosis, treatment planning, and prevention of complications.[1] However, despite the advances in imaging resources, the diagnosis of some root alterations such as vertical root fractures (VRFs), external root resorptions (ERRs), and root perforations (RPs) may be extremely challenging, especially when specific clinical symptoms and signs are not present.[23]

VRFs are characterized by the presence of a complete or incomplete fracture line along the long axis of the tooth.[4] Early detection of VRFs is fundamental for the therapeutic strategy. The prognosis of VRF cases when the root dentin is involved is poor, which may result in tooth extraction.[56] Although the direct visualization of a radiolucent line in radiographs is the only explicit characteristic of VRFs,[7] its detection may be difficult.[78]

ERR is an irreversible multifactorial destruction process of the dental structure, which may also lead to tooth loss.[9] In general, it appears on radiographs as root shortening or defects on the root surface.[10] During the initial stages, this alteration is asymptomatic and may not be detectable in periapical radiographs.[1112] RPs, on the other hand, are characterized by the presence of a communication between the root canal and outer surface of the tooth. It is a clinical accident that may occur at any stage of endodontic treatment and is usually iatrogenically induced. Most perforations tend to occur in the mandibular molars due to root curvature.[13]

Different factors such as milliamperage, field of view, voxel size (volume elements), detector type employed for image acquisition, device design, and patient stability may all influence cone-beam computed tomography (CBCT) quality and result in different image resolutions.[141516] In general, the smaller the voxel size and the longer the scanning time, the better the image resolution.[17] However, the level of radiation the patient is submitted during CBCT must always be taken into consideration. Ideally, a balance between obtaining the best possible images and exposing the patient to the least radiation levels possible should be reached.

Therefore, the objective of this in vitro study was to assess the accuracy of CBCT images of teeth with VRFs, ERRs, and RPs acquired with five different voxel sizes.

MATERIALS AND METHODS

The study was conducted in accordance with the Declaration of Helsinki and initiated only after the approval from the Institutional Review Board.

Sample preparation

Forty human premolar teeth were randomly allocated (drawing lots) into four experimental groups (n = 10): Sound teeth, reserved without any alterations (control); ERR; RP; and VRF. All simulated root alterations were performed by only one calibrated operator.

To produce external root alterations, the surface of the 10 teeth in the ERR group was ground with FG #1/4 round carbide burs (kG Sorensen, Cotia, Brazil) at high speed under constant water cooling. One alteration was performed perpendicularly to the surface of the middle third of the root. The teeth allocated in the RP group were intentionally perforated with FG 1011 HL diamond burs (kG Sorensen, Cotia, Brazil), mounted into a high-speed handpiece under water cooling at an angle of 45° in relation to the long axis of the teeth.

All the 10 teeth in the VRF group were initially submitted to inspection in a stereomicroscope SZX7 (Olympus, São Paulo, São Paulo, Brazil) to ensure that they were free from any previous complete or incomplete fracture as well as cracks or gaps. Dental crowns were sectioned 2 mm above the cementoenamel junction.[118] The teeth were embedded in hard PVC tubes measuring 20 mm × 25 mm × 25 mm (Grupo Tigre S/A, Joinville, Brazil), containing thermo-polymerizing acrylic resin (Clássico, São Paulo, Brazil). Embedded teeth were assembled inside a metal device and fixed to a universal assay machine (EMIC – DL 1000, São José dos Pinhais, Paraná, Brazil), in such a way as to receive vertical compressive force on the nucleus. The load was applied with the use of a spherical tip at a constant speed of 1 mm/min until fracture.[10] Maximum rupture strength, registered with Tesc® software, version 3.05 (Mattest Automação e Informática Ltda, Poá, Brazil), varied from 435.47 N to 472.68 N. No fragment dislodging occurred during the production of vertical fractures.

Cone-beam computed tomography image acquisition

After preparation, forty experimental teeth were assembled inside the sockets of ten macerated human mandibles. Pink wax No. 7 (Polidental, Cotia, Brazil) was used to keep teeth in position. Each mandible was prepared with four randomly distributed teeth, regardless of the group they belonged. Mandibles with the teeth were then positioned with the occlusal plane parallel to the ground and the median–sagittal plane perpendicular to the ground and then submitted to CBCT (i-Cat Next Generation®, Imaging Sciences International, Hatfield, Pennsylvania, USA). Images of each mandible were acquired with five different voxel sizes. CBCT settings are shown in Table 1.

Table 1

Cone-beam computed tomography (i-Cat Next Generation®) setting for image acquisition of root alterations

Cone-beam computed tomography scan analysis

All images generated were stored in a computer with an Intel® Core™ i7-2600 (Intel Corporation, Mountain View, California, USA) processor, a NVIDIA GeForce 9800 GT (NVIDIA Corporation, Santa Clara, California, USA) video board – and visualized with the iCATVision image software, version 1.8.1.10 (Imaging Sciences International, Hatfield, PA, USA), which comes with CBCT device. Panoramic, axial, and three-dimensional reconstructions, as well as the parasagittal slices, were visualized on a LED widescreen Dell monitor (16:09), S series, U2440L, 24 inches, maximum resolution of 1.920 × 1.080, intensity of 16.7 million colors, 86% color range (typical), and contrast dynamic rate of 8.000.000:1.

The scans were independently assessed by three radiology and dental imaging specialist examiners with at least 5 years of experience with CBCT, who were blind to the root alterations and also to the voxel sizes of acquisition [Figures 1 and 2]. The examiners were allowed to adjust image contrast, brightness, and magnification, and no specific time was established for the interpretation sessions. Root alterations were assessed using a confidence scale according to the following scores: (1) alteration definitely present, (2) alteration probably present, (3) unsure if alteration was present or absent, (4) alteration probably absent, and (5) alteration definitely absent; and 5 - alteration definitely absent. To determine inter- and intra-examiner reliability, all images were assessed in two different moments with an interval of 2 weeks between the first and the second assessment.[1610]

Figure 1

Axial, panoramic, and three-dimensional reconstructions, and parasagittal cuts, with voxel size resolution of 0.125 mm, of a sound tooth (iCATVision image software)

Figure 2

Parasagittal slice images, voxel size resolution of 0.125 mm obtained with cone-beam computed tomography (i-Cat Next Generation® and iCATVision image software): (a) external root resorption; (b) vertical root fracture; and (c) root perforation

Data analysis

The minimum sample size was calculated with the power of the test set at 95% according to the equation proposed by Lwanga and Lemeshow.[19] Inter- and intra-examiner agreement was conducted with Cohen's kappa test. The areas under the curves were compared with the Tukey exact binomial test (α=5%). All the tests were performed with SPSS®, version 22.0 (IBM Corporation, Armonk, NY, USA) for Microsoft Windows software at a level of significance of 95% (P ≤ 0.05).

RESULTS

The minimum number of teeth calculated for each group was seven. Interexaminer reliability was considered excellent for the method studied whereas the intraexaminer examination revealed that Examiners 1 and 2 presented excellent agreement between assessments while Examiner 3 presented good agreement.

Data obtained for sensitivity, specificity, PPV, NPV, for each voxel size used in the assessment of the three different root alterations studied, are shown in Table 2. It can be observed that all roots’ alterations were identified correctly (maximum sensitivity values) with voxel sizes 0.125, 0.200, and 0.250 mm. Voxel sizes 0.300 and 0.400 mm also presented high sensitivity.

Table 2

Diagnostic assessment according to different voxel sizes for teeth with root resorption, root fracture, and root perforation

The receiver operating characteristic curves obtained with the five different voxel sizes evaluated for each root alteration are shown in Figure 3. When the accuracy of images obtained with the five different voxel sizes used was compared, no statistically significant differences among voxel sizes 0.125, 0.200, and 0.250 mm were found in the identification of ERRs and VRFs. Although voxel sizes 0.300 and 0.400 mm presented no statistical differences between them, they performed statistically worse than the other three smaller voxel sizes. No statistically significant differences were found among any of the voxel sizes studied in the identification of RPs. Accuracy values with voxel sizes 0.125 and 0.250 mm were, in general, lower whereas with voxel sizes 0.300 and 0.400 mm were higher for RPs than those found for ERRs and VRFs.

Figure 3

Receiver operating characteristic curves for the three root alterations studied: (a) root resorption; (b) vertical root fracture; and (c) root perforation

DISCUSSION

According to the As Low As Reasonably Achievable (ALARA) principle, introduced by the International Commission on Radiation Protection in 1990,[20] it is important to select methods of diagnosis that use the least, reasonably practicable radiation doses to obtain images of adequate quality to reach a precise diagnosis.[1621] Nonetheless, if the clinical characteristics of the patient, together with the conventional radiographic data, are insufficient to provide the required information, CBCT may be indicated.[212223] The objective of this study was to analyze the accuracy of CBCT images obtained with different voxel sizes in the identification of different root alterations, to recommend the best protocol in compliance with the ALARA principle.

Vier-Pelisser et al.[24] revealed that ERRs might be present without being radiographically visible. In the present study, voxel size 0.250 mm was enough to ensure 100% accuracy (100% correction in determining both the presence and absence of an alteration) in the identification of ERR, dispensing with the need for smaller voxel sizes. The worst sensitivity and specificity results in the identification of ERRs were obtained with voxel size 0.400 mm, followed by voxel size 0.300 mm. Liedke et al.,[9] when assessing the diagnostic capability of CBCT with different voxel sizes in detecting simulated ERRs, suggested that adequate diagnostic performance with the least radiation could be obtained with voxel size 0.300 mm, even though voxel size 0.200 mm had provided better accuracy results. Under the conditions of this study, although sensitivity with voxel size 0.300 mm was high (97%), specificity was relatively low (77%), resulting in 87% accuracy. A result could compromise the clinical assessment of ERRs.

In the case of VRFs, voxel size 0.200 mm was enough to produce 100% accuracy while voxel size 0.250 mm resulted in 90% accuracy. For this particular type of root alteration, voxel sizes 0.300 mm and 0.400 mm provided relatively poor accuracy when compared to the other root alterations studied. This seems to indicate that images with these voxel sizes may seriously compromise operator's ability to correctly identify VRFs. These results are in agreement with Özer,[16] who demonstrated better precision in the detection of VRFs with voxel size 0.125 mm, followed by 0.200 mm. According to the author, although larger voxel sizes (0.300 and 0.400 mm) presented less truthful images, they were considered acceptable under clinical conditions. Junqueira et al.[25] also used i-Cat Next Generation® to detect fractures in teeth with and without intracanal posts. The authors found that in teeth without intracanal posts, voxel size 0.125 mm presented 99% accuracy while with voxel size 0.250 mm accuracy fell to 79%. In the present study, however, the accuracy with voxel size 0.250 mm was higher (95%), a difference that may be explained by differences in study design.

In the case of teeth with RPs, the best accuracy obtained was 90%, with no differences among voxel sizes 0.125, 0.200, and 0.250 mm. Although sensitivity values with voxel sizes 0.300 and 0.400 mm were quite high, specificity values were low, reducing accuracy in the analysis of images. However, the lack of statistical difference among the voxel sizes tested indicates that RPs may be difficult to identify, regardless of the voxel size. In a study conducted by Shemesh et al.,[12] sensitivity and specificity of deep perforations were 86% and 70%, respectively, reinforcing the perception that it not always possible to obtain clear images of RPs, even with CBCT.[12]

Although all participating examiners were experienced in the analysis of CBCT images, the intraexaminer reliability test demonstrated that Examiner 1 was more consistent than Examiner 2, who in turn was more consistent than Examiner 3 in the assessment of the three root alterations studied. This clearly demonstrates the important role of examiner's experience and ability to assess CBCT images, regardless of the settings of the equipment (i-Cat Next Generation®). Thus, taking into consideration the conditions this study was conducted, it would seem logical to suggest that the smallest voxel size (0.125 mm) should be used to assess all the three root alterations studied. Even though voxel sizes 0.250 and 0.200 mm would be enough to guarantee 100% accuracy in the analysis of ERRs and VRFs, respectively, voxel size 0.125 mm would offer the best possible image resolution without increasing the amount of radiation to the patient. In the case of RPs, the lack of statistical differences among the voxel sizes tested seems to suggest that even voxel size 0.400 mm would be clinically acceptable. However, voxel size 0.125 mm would, again, ensure the best resolution possible and, therefore, the best possibility of a precise diagnosis.

CONCLUSIONS

Voxel size 0.125 mm produced images with the best resolution with no increase in radiation levels to the patient when compared to voxel sizes 0.200 and 0.250 mm. Voxel sizes 0.300 and 0.400 mm should be avoided in the identification of root alterations.

Financial support and sponsorship

Nil.

Conflicts of interest

There are no conflicts of interest.