The OCT system (Prototype, LG Electronics, Seoul, Korea) developed for this study was based on the swept-source method (Figure 1). The swept source (AXP50124-8, Axsun Technologies, Boston, MA, USA) generates light at a center wavelength of 1,310 nm, with a repetition rate of 50 kHz, an average power of 10 mW, and a sweep range of 110 nm. The emitted beam is split in a directional coupler into 2 arms, reference and sample, with 90% of the beam passing through the sampler arm and 10% through the reference arm. The circulator in each arm directs these beams toward identical collimators. Then, in the sample arm, a 2D galvanometer system steers the beam and focuses it into the sample tissue via a scan lens. The scan lens has a numerical aperture of 0.026, resulting in a spot diameter of approximately 32 μm in air. Meanwhile, in the reference arm, the beam passes through a dispersion compensation block—to match the dispersion introduced by the scan lens—and is then reflected by a mirror. Both reflected beams are routed via the circulators toward another directional coupler, which evenly splits the 2 incoming beams into 2 ports. Each port transmitting the halves of the beam reflected from the sample and the reference mirror is connected to a balanced detector. This detector subtracts the signal from 1 port from that of the other to remove the common signal and leave only the interference signal. The polarization controllers in both arms are used to maximize the interference signal. The resulting signal is then processed on a PC to generate a depth-resolved intensity profile of the light backscattered from the tissue under investigation. This process is repeated whenever the galvanometer changes its angle in order to generate 2D or 3D images. The system has an imaging capability of capturing 25 frames per second with an axial resolution in air and water of 20 and 30 µm, respectively, and a lateral resolution of 32 µm. The 2D images present 500 pixels per column on the x-axis, corresponding to a maximum of 10 mm of scanning, and 512 pixels per column on the y-axis, corresponding to a maximum depth of 5 mm.

Animal selection and management, as well as the surgical protocol and preparation used in this study, followed protocols approved by the Institutional Animal Care and Use Committee, Yonsei Medical Center, Seoul, Korea (Permission No. 2011-0188). Five beagle dogs were used to produce the ex vivo specimens. Ten block-bone samples containing the second premolars from each side of the mandible were collected, and the specimens were preserved in 10% formalin prior to the experiments.

OCT images of the gingival sulcus and adjacent root surface were captured at the midbuccal aspect of the second mandibular premolars of the 10 specimens. A single blinded examiner measured the biologic depth in each captured image using the OCT software interface. The biologic depth was measured as the distance from the gingival margin to the base of the sulcus, as visible on each scanned image (Figure 2). In more detail, a combined vector (the biologic depth) was deduced from the horizontal and vertical measurements on the x- and y-axes of the OCT interface. The presence of dental calculus in the captured images was assessed based on the visible signal intensities.

Statistical analysis was performed with SPSS version 23 (IBM Corp., Armonk, NY, USA). To evaluate the measurement error, the intraclass correlation coefficient between the OCT and histologic measurements was calculated in a 2-way random model. In addition, the Bland-Altman method was used to evaluate the 2 techniques. In the Bland-Altman method, to determine whether there was a level of agreement between the 2 techniques or a proportional bias, the 1-sample t-test was used, followed by a visual inspection of the Bland-Altman plot, and then a linear regression analysis.

All harvested specimens were scanned with a micro-CT system (Sky-Scan 1173, SkyScan, Aartselaar, Belgium) at a resolution of 13.85 μm (achieved using 130 kV and 60 μA). The scanned data set was processed in Digital Imaging and Communications in Medicine (DICOM) format, and cross-sectional views and 3D images of the specimens were reconstructed with On-Demand 3D software (Cybermed, Seoul, Korea).

Block sections of the surgical sites were fixed in 10% formalin for 10 days. The fixed specimens were decalcified in 5% formic acid for 14 days and then embedded in paraffin. Serial 5-μm-thick sections were cut buccolingually at the central portion of the second premolars. The midbuccal sections identical to the sites where measurements were made using OCT were selected and stained with hematoxylin and eosin for histologic observations. The histologic slides were observed and digitally captured under a light microscope (DM LB, Leica Microsystems, Wetzlar, Germany) equipped with a camera (BX50, Olympus, Tokyo, Japan). After examination using conventional microscopy, computer-assisted histometric measurements of periodontal dimensions, including the sulcus depth, gingival thickness, and epithelial and connective tissue thickness, were made using an automated image analysis system (Image-Pro Plus, Media Cybernetics, Silver Spring, MD, USA).

Periodontal anatomical structures were imaged using ex vivo dental OCT (Figure 2). The periodontal tissue contour, gingival sulcus, and the presence of supragingival and subgingival calculus could be visualized in high definition using the OCT system. The surface microstructural characteristics of both the tooth and gingiva could be assessed. The dentinoenamel junction (DEJ) could be seen in the tooth images. The deposition of dental calculus on tooth surfaces both above and below the gingival margin could be visualized, and this was confirmed in a graph produced using the software interface showing increased signal amplitude compared to the neighboring tooth surface (Figure 3). Calculus deposits or anomalous tooth surfaces presented as sharp peaks in signal amplitude compared to clean sites (Figure 4), and this finding was consistent in both supragingival and subgingival calculus specimens.

Depth measurements could be made using the software interface either in real time or on previously captured images. The epithelial layer and the subepithelial connective tissue layer were distinguishable in the images as different signal intensities (Figure 5A). The epithelial layer of the gingiva was measured at approximately 0.5 mm, which was consistent with the histologic measurements. However, the overall imaging depth was typically 1.2–1.5 mm, which meant that the presence of alveolar bone could not be visualized on the images.

The depth of the gingival sulcus as measured using OCT and histologically was compared (Table 1). This was performed at the midbuccal aspect of 10 mandibular second premolars. The OCT and histologic measurements showed an intraclass correlation coefficient of 0.220 (95% confidence interval: −2.141, 0.806), which suggests low correspondence between the two measurement methods. The Bland-Altman method revealed a significant level of disagreement between the 2 methods (P=0.008). Inspection of the Bland-Altman plot (Figure 6) showed that the differences in the measurements were greater at deeper biologic depths; however, statistically, no significant proportional bias was found (P=0.101). The mean difference between 2 methods for measurements at the 10 sites was 0.41 mm, with the OCT measurements being shallower (1.20±0.19 mm) than the histologic measurements (1.61±0.35 mm).

The micro-CT images could effectively display the teeth and surrounding alveolar bone in 3D reconstruction views as well as in cross-sectional views (Figure 7). In the cross-sectional view, the dense cortical outline of the alveolar bone could be seen, bundle bone surrounded the roots of the teeth, and cancellous bone with an interconnected trabecular pattern was evident in detail against the radiolucent bone marrow (Figure 7B) [10]. The internal structures of the teeth could also be observed; the presence of the DEJ was clearly depicted due to the high radiopacity of the enamel. Such details were not visible on OCT images due to the poor imaging penetration.

Soft-tissue contours could be made visible by appropriately adjusting the grayscale used to display the images. The gingival outline including the location of the vestibule could be seen, and the thickness of the soft tissue could be measured from the bone in cross-sectional views. However, certain aspects of the microstructural architecture, such as the gingival sulcus and connective tissue attachments, could not be distinguished.

Histologic sections were used as the gold standard in comparisons with OCT and micro-CT images (Figure 5C). The process of embedding paraffin wax into the gingival sulcus had expanded the space inside the sulcus, and so the histologic sections appeared different from the ex vivo OCT images. However, the overall shape of the tooth and the surrounding soft-tissue contours remained very similar. The DEJ was evident only in some sections, and in those, the enamel appeared to be thinner. This may have been caused by the dissolution of enamel during the decalcification process using formic acid. Orthokeratinized gingival epithelium formed long rete pegs against the underlying connective tissue, unlike the junctional epithelium, which appeared to be only a few cells thick, flat, and without rete pegs (Figure 5C). The histologic slides were obtained from the coronal section of the furcation area. The furcation was chosen deliberately since it is frequently associated with the progression of periodontal disease. However, despite the natural accumulation of calculus at some sites, no significant attachment loss resulting in furcation involvement was detected.