Assessment of thin bony structures using cone-beam computed tomography

Assessment of thin bony structures with cone-beam computed tomography (CBCT) has attracted great interest in various disciplines since its introduction in the late 1990s. In orthodontics, it is an essential part of studying the possible side effects of the marginal bone level due to treatment and retention.^1,2  To achieve accurate results, the technique must be properly used. The vast majority of studies investigating the validity and reliability of bone level measurements in relation to thin bony structures on CBCT images have focused on the inherent technical properties of the CBCT unit (eg, voxel size), field of view (FOV), and the exposure parameters (e.g., milliampere [mA], and kilovolt [kV]).^3–5  Other parameters, such as spatial resolution and partial volume averaging, may also play an important role in the visualization of small or thin structures.

Crucial to achieving high validity and reliability, however, is that the raw radiologic data are reconstructed in a way that optimizes visualization. Usually, this is done using multiplanar reconstruction (MPR), which produces three image planes orthogonal to each other (axial, sagittal, and coronal) to depict image volume; this enables the observer to choose the best view for the diagnostic task. Another option is to use three-dimensional (3D) volume rendering, which, in comparison with MPR, presents the radiological data as a 3D surface object. Besides these two reconstruction techniques, the observer can vary the viewing mode for brightness, contrast, and pixel value (MPR) and enhance or reduce the visualization of tissues with different densities (3D).

The effects of various technical and exposure parameters on the visualization of thin bony structures, expressed as the validity and reliability of marginal bone level measurements, have been thoroughly investigated. However, the effects of various reconstruction techniques and viewing modes have garnered less interest.

In an attempt to optimize the assessment of thin bony structures on CBCT images, the present study evaluated the validity and reliability of marginal bone level measurements made on CBCT images produced using two reconstruction techniques, two viewing modes, and two resolutions and compared them with the gold standard of histologic measurements.

The material consisted of 16 anterior mandibular teeth (incisors and canines) from six human specimens. The Ethics Review Committee at Uppsala granted ethical approval (Daybook no. [Dnr]: 2019-01582) for the study protocol.

Prior to the radiographic examination, each tooth was assigned a unique identification code, and a metal ligature was attached to the buccal aspect of the tooth crown along the long axis. The ligature served as a marker to ensure correct positioning during radiographic measurements in relation to the histologic section.

Before radiographic examination, each specimen was surrounded by a tissue-equivalent material, imitating soft tissue (Superflab, Eckert & Ziegler, BEBIG GmbH, Berlin, Germany). The Accuitomo 170 3D CBCT unit (J. Morita Mfg. Corp, Kyoto, Japan) was then used to image the tooth. Exposure parameters were 75 kV, 5 mA, and an FOV of 60 × 60 mm. A full 360° rotation was made during imaging. Each tooth was fully imaged twice: once at standard resolution (121 images) and once at high a resolution (481 images). All image data were exported in the Digital Imaging and Communications in Medicine (DICOM) format with a slice thickness of 0.5 mm (standard resolution) or 0.125 mm (high resolution) for later import and viewing in OsiriX MD, a medical image viewer for viewing DICOM images (Pixmeo SARL, Bernex, Switzerland).

The first step in preparing the raw data for analysis was to make the MPRs in the axial, coronal, and sagittal viewing planes. The planes were oriented in relation to the long axis of the tooth and the metal ligature (Figure 1). Image reconstruction enables an optimized visualization of the tooth and surrounding marginal bone. Measurements were made between the cemento-enamel junction and the marginal bone level. On MPR images, measurements were made at the buccal and lingual aspects of each tooth with two viewing modes: gray scale and inverted gray scale. On the 3D-rendered images, measurements were made after the image had been optimized for visualizing bone (Figure 2). A specialist in dentomaxillofacial radiology made all measurements.

The teeth were preinfiltrated with a mixture of ethanol and Technovit® 7200 VLC (glycol methacrylate, 2-hydroxyethyl methacrylate) in decreasing (ethanol) and increasing (Technovit) concentrations in three steps (70/30, 50/50, and 30/70) followed by 2 weeks of infiltration in pure Technovit 7200 that was changed at halftime. The Technovit was cured in ultraviolet and white light overnight to achieve polymerization. The teeth were then prepared according to Donath⁶ : they were cut along the long axis using a metal wire as a guide before grinding to a thickness of about 100 μm. No further staining was done. A light microscope (Nikon SMZ 800, Nikon, Tokyo, Japan) connected to a Nikon DS-Fi1 camera and a computer with NIS-Elements Documentation software from Nikon were used to photograph tooth slices; measurements were made with a 0.5× lens (1–6.4×) and a magnification of 10× (0.5 × 1 × 10).

One histological section (one tooth) was excluded due to displacement from the alveolus during histologic preparation; thus, the final sample comprised 15 teeth.

The histomorphometric evaluations were made by one observer (Dr Westerlund).

Histomorphometric evaluations were repeated at intervals of at least 10 days in order to measure intraobserver reliability. To measure inter- and intraobserver reliability of radiographic evaluations, measurements were repeated at intervals of at least 10 days on a randomized selection comprising 38% (one tooth/histological specimen) by two specialists in dentomaxillofacial radiology (Ms Lennholm, Dr Lund).

The intraclass correlation coefficient (ICC) and descriptive statistics (mean, standard deviation) were used to calculate the inter- and intraobserver reliability. Mean differences between radiologic (CBCT) and histologic measurements were used to calculate validity. Bland-Altman plots were used to depict agreement. Wilcoxon signed rank tests were used to compare paired samples. The level of statistical significance was set at P < .05. The Statistical Package for the Social Sciences (IBM SPSS, version 27.0; IBM Inc, Chicago, Ill) or GraphPad Prism (version 8.3.1; GraphPad Software, San Diego, Calif) were used for all statistical analyses.

The ICC was 0.99 (95% confidence interval [CI]: 0.98–1.00) for buccal and lingual measurements. The mean difference between intraobserver measurements 1 and 2 was 0.04 mm (95% CI: −0.11 to 0.19 mm) for buccal measurements and 0.03 mm (95% CI: −0.08 to 0.14 mm) for lingual measurements.

Intraobserver variation was assessed for the standard-resolution and high-resolution images using the two viewing modes (MPR) and 3D rendering. The measurements of observer 1 varied between ICCs of 0.94 and 1.00 (95% CI: 0.63–1.00) except for one outlier (standard resolution, 3D rendering, buccal aspect; ICC 0.50 [95% CI: −1.19 to 0.92]). The measurements of observer 2 varied between ICCs of 0.86 and 1.00 (95% CI: 0.43–1.00).

The interobserver variation between observers 1 and 2 varied between ICCs of 0.74 and 0.97 (95% CI: 0.12–0.99).

The mean differences and standard deviations for intra- and interobserver assessments are shown in Table 1.

Comparisons between the CBCT and histological measurements were made at the buccal and lingual surfaces, on standard- and high-resolution protocols, using MPRs with two viewing modes and 3D-rendered images.

Comparisons of radiologic (CBCT) and histologic measurements yielded the highest validity at the buccal surfaces using the standard protocol, MPR, and the inverted gray scale viewing mode (mean difference = 0.02 mm). The lowest validity occurred at the lingual surfaces using the high-resolution protocol and 3D-rendered images (mean difference = 1.10 mm). Bland-Altman plots (Figure 3a,b) depict the agreement between the two MPR viewing modes (gray scale [gs], inverted [inv]), 3D rendering, and histology. At the lingual surfaces, the mean differences with both resolutions were significant (P < .05) for both viewing modes (MPR) and 3D rendering. At the buccal surfaces, the mean differences were significant (P < .05) when using high-resolution and 3D rendering.

The assessment of thin bony structures is an essential component of radiologic diagnosis and has gained widespread interest in various disciplines, including orthodontics. Previous research, however, has focused mainly on technical properties and exposure parameters. The present study strove to assess the next link in the imaging chain, namely, the effect of a reconstruction technique and of two viewing modes on the observer's ability to detect thin bony structures.

In general, the present study found that 3D rendering and two viewing modes with MPR images had little influence on visualization of thin bony structures. Although at the lingual surfaces, the limits of agreement were wider and thus the variation larger than at the buccal surfaces despite protocol, reconstruction, and viewing mode used. Technical properties and exposure parameters, such as the resolution used in image acquisition, are probably more important.

In comparisons with the histologic reference, 3D-rendered images had somewhat lower validity at the lingual surfaces than when MPR images were presented in the gray scale or inverted gray scale viewing mode. These findings agreed with those of Fernandes et al.,⁷  in that measurements on the MPR images with commonly used gray scale settings were more accurate than measurements on 3D-reconstructed images. CBCT, in contrast to conventional computed tomography (CT), does not use calibrated Hounsfield units. Thus, the ability to distinguish between tissues of differing densities is more difficult on CBCT than on CT images; therefore, when trying to optimize a 3D-rendered image by varying screen settings, alterations may occur in the visualization of the outer surface and, as a consequence, of the remaining amount of bone. Bone tends to be thinner on the lingual surfaces of the alveolar process, which affects the reliability of the assessment; thus, the greater the variation in radiologic measurements, the greater the difference in the histological reference and, consequently, the lower the validity. In the present study, this was evident not only on the 3D-rendered images but also for MPR images visualized using a gray scale and inverted gray scale viewing mode. Results for these three situations were similar, with differences being significantly higher at the lingual aspects.

Reliability of bone level measurements at the buccal aspects of the teeth were higher than at the lingual aspects, as was validity, where differences between radiologic measurements on the CBCT images and of the histologic reference were lower. Neither reconstruction technique (3D or MPR) nor any MPR viewing mode (gray scale or inverted gray scale) showed improved bone level measurements of clinical relevance compared with the histologic reference. Measurement reliability and thus validity at the buccal aspects of the teeth on 3D-rendered images was better than on MPR images presented in the gray scale or the inverted gray scale viewing mode; the narrower limit of agreement in the Bland-Altman plot illustrates this. A probable explanation is that, in the 3D-rendered images, the outer surface of the bone and thus the remaining amount of bone, may be altered and “smoother,” yielding less variation in visualization of the marginal bone.

To discern the influence of technical properties and exposure parameters on CBCT visualization, the present study investigated two reconstruction techniques, two viewing modes, and two acquisition resolutions. Other parameters such as FOV, kV, and mA were kept constant at exposure. This was done to compare the day-to-day clinical setup with a proposed optimal setup. Sun et al.³  investigated the accuracy of alveolar bone-height measurements on CBCT (I-Cat) images using different technical parameters (voxel sizes) and specimens with varying bone thicknesses; they found that measurement accuracy increased with smaller voxel size. The present study used a CBCT unit with a voxel size of 0.125 mm. A previous study investigating the influence of different exposure parameters, such as varying FOVs, mAs, and kVs, and with and without a copper filter, concluded that the presence or absence of a filter and variations in kV settings did not affect overall image quality, although the higher mA setting was always preferred.⁴  Better image quality is associated with more photons per voxel, and the mA settings determine the amount of photons. Other studies, such as Dillenseger et al.,⁵  presented contrasting results concerning the influence of different FOVs; they found that the quality of images made using a small FOV was similar to the quality of those made with a large FOV. Molen, however, found that images made with larger FOVs had greater noise from scatter and worse spatial resolution.⁸  Images made with smaller FOVs do have lower scatter noise, but spatial resolution is poorer due to a higher sensitivity to noise. This is because, given the same exposure parameters, voxel size is smaller with smaller FOVs; hence, to achieve a similar level of noise, exposure parameters need to change accordingly.

Concerning the effect of imaging protocol on image quality, the present study could not confirm previous findings that the validity of high-resolution protocols for detecting marginal bone level was higher than of lower-resolution protocols.^9,10  In agreement with Ruetters et al.,¹⁰  the difference between the two protocols in the present study was small and of limited clinical relevance. Thus, the approximately 40% higher radiation dose delivered by the higher-resolution protocol is considered unjustifiable.¹¹ 

The results of the present study were comparable with previous studies investigating the validity of CBCT for assessing marginal bone level with a focus on the technical properties of the system and exposure parameters.^12–14  However, as Thönissen et al.¹⁵  stated, CBCT imaging is unreliable for visualizing thin bony structures less than 0.5 mm in width even if high resolution is used. Apart from the various technical parameters of the system, other factors may play an important role in the visualization of small or thin structures, such as spatial resolution and partial volume averaging. Spatial resolution is the minimum distance needed between two objects to distinguish between them. Partial volume averaging is the result of densities between adjacent voxels being combined, since each voxel can display only one grey level at a time; thus, for example, when thin cortical bones adjacent to air share a voxel, the voxel displays a median density. So, the gray scale bit depth of the system also plays an important role in assessment. Today, most CBCT systems use a bit depth between 12 and 16, corresponding to 4096 (2¹²) and 65,536 (2¹⁶) shades of gray, respectively.¹⁶ 

The present study used human specimens from the anterior mandibular region of the mouth, as this area in general has a smaller buccolingual width, and thus, the teeth are covered with thinner bone that may be further affected if the teeth move closer to the buccal or lingual plate, for example, during orthodontic treatment.^17,18  The rationale for using human specimens was to achieve a situation as close as possible to the clinical situation. However, compared with an in vivo study, the setting is not prone to image artifacts due to movement nor artifacts from metallic materials, which could occur in clinical conditions, influencing assessment.

When evaluating thin bony structures and marginal bone level in orthodontics:

• 3D rendering and MPR images with an inverted gray scale viewing mode do not improve the observer's ability to visualize thin bony structures in the anterior mandibular region.

• The use of 3D-reconstructed CBCT images to assess thin bony structures should be avoided when thin cortical borders, such as in the anterior mandibular region, are suspected.

• The small difference between imaging protocols (standard and high-resolution) in visualization of thin bony structures does not justify the use of the higher radiation dose required using the high-resolution protocol; thus, no changes in current, clinically used protocols (standard resolution) are recommended at this time.

The authors want to express their sincere gratitude to Petra H. Johansson, Department of Prosthodontics/Dental Material Science, Institute of Odontology, Sahlgrenska Academy, Gothenburg University, Gothenburg, Sweden, for excellent histologic preparation of the specimens.