2.1.

Specification of the Rigid Endoscopic Optics

In this research, the rigid endoscopic optics used for OCT imaging of the oral cavity is a commercially available near-infrared (NIR) endoscope from Karl Storz GmbH & Co. KG. The NIR endoscope has an insertion diameter of 10 mm, a length of 200 mm, and an angle of view of 0 deg (instrument axis consistent with optical axis) and is assumed to be based on afocal optics with a 0.25-fold magnification. For a center wavelength of ${\lambda }_{\text{center}}=580\mathrm{nm}$, a lateral resolution of $\mathrm{\Delta }x=62.5\mu \mathrm{m}$ at a working distance of WD $\sim 49\mathrm{mm}$ is determined by means of an externally illuminated USAF resolution test target (USAF-1951-target) in combination with a proximally positioned VIS camera focused to infinity. In order to get a focus on the distal side of the endoscope and to increase the lateral resolution, the optics in front of the proximal end of the endoscope was simply adapted. As a result, a lateral resolution of $\mathrm{\Delta }x=12.4\mu \mathrm{m}$ at WD $\sim 7.5\mathrm{mm}$ and ${\lambda }_{\text{center}}=580\mathrm{nm}$ is achieved [cf. Fig. 1(b)], which corresponds to $\mathrm{\Delta }x=17.9\mu \mathrm{m}$ for the NIR wavelength range with ${\lambda }_{\text{center}}=835\mathrm{nm}$ under the assumption of $\mathrm{\Delta }x\sim {\lambda }_{\text{center}}$. The FoV in dependency on WD is determined by a millimeter paper test chart with the result of a linear relationship with a slope of $m=1.1$, as shown in Fig. 2(c).

Fig. 1

By adapting the proximal optics for light coupling, a lateral resolution of $\mathrm{\Delta }x=12.4\mu \mathrm{m}$ of the commercial NIR endoscope from Karl Storz GmbH & Co. KG is measured by means of a proximal VIS camera in combination with the USAF resolution test target, which is positioned at a working distance of WD $\sim 7.5\mathrm{mm}$ and illuminated by a center wavelength of ${\lambda }_{\text{center}}=580\mathrm{nm}$.

Fig. 2

(a,b) Endoscopic VIS images of a millimeter paper test chart, which is illuminated by the integrated illumination system of the endoscope, for measuring the relation between the FoV and the WD. (c) At the favorable WD of 5 to 10 mm, the FoV with fan-shaped beam increased linearly from 8.3 to 14.5 mm.

2.2.

Endoscopic Optical Coherence Tomography System Setup

The OCT system used in this study is spectrometer-based and consists of a fiber-coupled superluminescent diode (SLD-371-HP1, Superlumdiodes Ltd., Ireland) with a central wavelength of 835 nm and a FWHM of 50 nm, an optical circulator (CIR-0850-40-APC, Thorlabs GmbH, Germany), a fiber-coupled Michelson interferometer and a self-developed spectrometer with a wavelength range from 802 to 867 nm. As presented in Fig. 3, the scanning EOCT probe is equipped with a collimator C1 (Schäfter + Kirchhoff GmbH, Germany), two galvanometer scanners ${\mathrm{GS}}_x$ and ${\mathrm{GS}}_y$ (Cambridge Technology Inc., USA) for 2-D beam deflection, a lens combination consisting of two achromats L1+L2 for light coupling (Edmund Optics GmbH, Germany), the NIR endoscope from Karl Storz GmbH & Co. KG, and an additional objective lens L3 (AC050-008-B, Thorlabs GmbH, Germany; at 4-mm distance to the distal end of the endoscope) for telecentric OCT imaging. As seen, the difference in glass content between sample and reference arm is not compensated optically but numerically in the processing.^68,69

Fig. 3

(a) Scheme of the principle OCT setup. SLD, 5-mW superluminescent diode with ${\lambda }_{\text{center}}=835\mathrm{nm}$ and $\mathrm{FWHM}=50\mathrm{nm}$; OC, optical circulator; FC, wideband fiber coupler with a coupling ratio of $75:25$; C1 and C2, collimators; GS, galvanometer scanners; L1–L4, lenses; endoscope, endoscope from Karl Storz GmbH and Co. KG; CLS, cold-light source; PC, polarization controller; RM, reference mirror; spectrometer, line rate of 11.88 kHz. (b) Telecentric imaging is realized by an additional lens L3 mounted on the distal end of the commercial endoscope at 4 mm distance by a custom adapter made of stainless steel.

The optical components C1, ${\mathrm{GS}}_x$, ${\mathrm{GS}}_y$, L1 and L2 are integrated in a base body made of polylactide (PLA), at which the chosen endoscope, a CCD VIS camera, and an ergonomic handle are mounted, as shown in Fig. 4(a). Lens L3 is pasted into an additional adapter made of stainless steel by two set screws (M2) and secured by optical glue (Vitralit^®1655, Panacol-Elosol GmbH, Germany). Since the distal lens allows imaging only in a small range, a CCD camera (acA1600-20uc, Basler AG) is installed at the base body of the handheld scanner head for online off-axis video guidance by means of an overview image of the oral cavity [cf. Fig. 4(a), yellow beam path], which is displayed on the integrated LCD screen (LCDInfo Oy, Finland). The custom-developed spectrometer, operating at a read-out rate of 11.88 kHz for one A-scan, is mounted in an OCT table device [cf. Fig. 4(b)] along with the SLD and the optical circulator, as well as the drive electronics of the galvanometer scanners and electronic control of the line scan detector (Dalsa IL C6, Teledyne DALSA Inc., Canada).⁷⁰ The control of the system, the data acquisition, and online display are controlled by custom software developed with LabVIEW^® (National Instruments Inc., USA) on a personal computer [cf. Fig. 4(c)]. With typical 480 A-scans per frame (B-scan), the system achieves $\sim 23\text{frames}/\mathrm{s}$, which can be processed and displayed in real time.

Fig. 4

Scheme of the measurement setup for the endoscopic investigation of the oral cavity equipped with (a) the EOCT probe with integrated scanning sample arm, (b) the base OCT device with built-in SLD, spectrometer, partial fiber optics and system electronics, and (c) a personal computer for the control and triggering of the electronic components as well as the data display and acquisition.

The overall dimensions of the developed scanner head including the proximal handle are $76\mathrm{mm}\times 442\mathrm{mm}\times 251\mathrm{mm}$ ($\text{width}\times \text{depth}\times \text{height}$) with an endoscopic part relevant for the mouth of a length of 200 mm and a diameter of 15 mm tapered to 10 mm at the distal end [cf. Fig. 15 in Appendix]. The height of the scanner body is mainly defined by the height of the handle (100 mm) and the height of the LCD screen (130 mm). With this design, most of the oral cavity is reached proven by two experienced dentists and one physician. Limitation of the developed endoscope might occur for the imaging of the buccal side and the interdental space of second and third molars of patients with reduced mouth opening. The endoscopic probe enables imaging in noncontact and contact mode by using an additional spacer (distance piece), which is simply mounted on the lens adapter of lens L3 before medical examination and acts as a disposable item for future clinical use. Advantageously, the endoscope, considered as an attachment to the scanner head, can be cleaned and disinfected (by immersing into 3% Sekusept^® aktiv solution, Ecolab Deutschland GmbH, Germany) in the daily hospital routine. If necessary, steam sterilization can be performed by disconnecting the commercial endoscope.

2.3.

EOCT Scanner Performance

In order to evaluate the performance of the developed EOCT probe with telecentric scanning, several measurements were performed for the quantification of the axial and lateral resolution $(\mathrm{\Delta }z,\mathrm{\Delta }x)$, the FoV, the sensitivity, and the spectrometer-induced depth-dependent sensitivity loss. The axial resolution of the OCT system without the endoscopic extension is experimentally measured to be $\mathrm{\Delta }z=11\mu \mathrm{m}$ in air and in good agreement with the theoretical value of $\mathrm{\Delta }z=10.8\mu \mathrm{m}$. The measurement range is 2.75 mm. Since the glass content in the sample arm is higher than in the reference arm, the occurring dispersion mismatch is numerically compensated for optimal resolution.^68,71 This is achieved by multiplying the spectral data (which is formerly linearized in k-space, zero-padded to $2^{12}\text{pixels}$ and spectrally shaped) with a compensating phase term ${\mathrm{e}}^{-i\phi (k)}$.^69,72 The empirically determined phase correction amounts to

Fig. 5

En face projection of the USAF-1951-target (a) as well as a millimeter-chart (b) detected by telecentric EOCT for determining the lateral resolution and the FoV. (c) Corresponding OCT cross-sectional image (B-scan) of the millimeter-chart endoscopically detected. Scale bar in the cross-sectional image (c) corresponds to $500\mu \mathrm{m}$ in air.

The sensitivity is obtained from a glass substrate surface ($-14\mathrm{dB}$ reflector) and amounts to $-88\mathrm{dB}$. Due to the decreasing resolvability of interference signals with increasing frequency on the line detector of the spectrometer, a signal loss of 8 dB (sensitivity-roll-off) between zero and maximum depth is determined.

3.1.

Dental Imaging

The most promising application fields for OCT in dentistry are (1) the detection of carious lesions and the characterization of (de)mineralization processes, (2) the detection of defects at dental restorations, and (3) the assessment of erosion and cracks. Particularly for the imaging of well-defined boundaries, such as the interfacial layer between a dental restoration and the natural tooth surface, OCT might be a beneficial tool for the dentist to assess internal defects.^74,75

For testing the developed EOCT system at dental hard tissue, two volunteers were examined (A: male 30 years, B: female 33 years). Prior to the OCT measurements, oral examination was made by an experienced dentist in order to select specific regions for imaging. Additionally, photographic overviews of the oral cavity were made for the allocation of the imaged regions. First, OCT measurements of the incisors of volunteer A are performed with the handheld scanner head in noncontact mode. In order to reduce movement artifacts for the imaging of the premolars and molars, the distal end of the endoscope is equipped with an attached distance piece, which is positioned at the region of interest of the tooth.

Figure 6 shows the photograph of the maxillary central incisors [Figs. 6(a) and 6(b)] as well as the OCT B-scans [Figs. 6(c)–6(e)] at the corresponding lines in the zoom view in Fig. 6(b). In Fig. 6(c), the proximal region of tooth 11 and tooth 21 was imaged [at the red line in Fig. 6(b)]. The EDJ can be recognized by a well-defined, low-scattering band between the enamel (E) and the dentin (D).^19,24,25 Conspicuously, the enamel of both tooth 11 and 21 appears as light-dark progression containing homogeneous bands of varying intensity, which is assumed to be caused by the birefringent properties. In former research, it was shown that enamel appears as layered structure in polarization-sensitive (PS)OCT.^76,77 Although the light source is not polarized, the system presented in this work is sensitive to birefringence.⁷⁸ Figure 6(d) shows a shadow artifact caused by a strong scattering microcrack in the upper region of the enamel [asterisk, cf. Refs. 31, 79, and 80, at the green line in Fig. 6(b)]. Due to increased scattering, higher intensities and a reduced penetration depth of the OCT signal can be observed at the gingiva in Fig. 6(e) [cf. Refs. 57, 75, and 81, at the blue line in Fig. 6(b)].

Fig. 6

(a) Photographic overview and (b) magnification of the maxillary central incisors with corresponding exemplary OCT B-scans of the proximal region between tooth 11 and 21 (c, red), the enamel (E), the EDJ and the dentin (D) of tooth 11 (d, green) with breakage (asterisk), and the intersection between enamel and gingiva (e, blue). The OCT measurement was performed in noncontact mode. Abbreviations: E, enamel; D, dentin; EDJ, enamel–dentin junction; G, gingiva; CR, composite restoration. Coordinate system in (d) is equivalent in image part (c) and (e). Scale bar in the OCT images (c–e) corresponds to $300\mu \mathrm{m}$. Indicated dynamic range is valid for all depicted OCT images.

The composite restoration (CR) at tooth 11 [cf. magenta frame in Fig. 6(b)] is characterized by a homogeneous scattering layer in the OCT B-scans, as presented in Figs. 7(a) and 7(b). Interfacial gaps (green arrow in a) or even small cracks (green arrow in b) are visible at different positions in the restoration, which was found by previous nonendoscopic ex vivo research results.^31,74,82,83 The 3-D representation reveals a roughened intersectional profile of the restoration (c,d). There, two projections were chosen: (c) with manually manipulated sectional view (artificial edges marked in bright green) to image the interfacial gap of the composite restoration in the three-dimensional context and (d) with intact and untreated surface of the imaged composite-enamel-passage.

Fig. 7

Cross-sectional images (a, b) and 3-D representation (c, d) of the composite restoration (CR) within the magenta frame in Fig. 6(b). Green arrows show gaps (a) or even small cracks (b) in the restoration. The dental surface of the 3-D display in panel (c) is manually manipulated in the postprocessing (artificial edges marked in bright green) to provide a more instructive view of the interfacial gap of the composite restoration. Again, the OCT measurement was performed in noncontact mode. Abbreviations: E, enamel; EDJ, enamel–dentin junction; CR, composite restoration. Scale bar in (a) and (b) corresponds to $300\mu \mathrm{m}$. Indicated dynamic range is valid for all depicted OCT images.

Beyond the OCT imaging at the maxillary central incisors, the first premolars of both volunteers were examined by the developed EOCT scanner head. The upper row in Fig. 8 [parts 8(a)–8(c)] represents the OCT B-scan and 3-D projection of the cervical area of tooth 24 in subject B, showing the exposed dentin at the transition to the gingiva. Furthermore, a small white spot can be identified in the reduced enamel layer without clear diagnosis. The OCT examination of the mineralization defect at tooth 24 in subject A can be recognized by a high scattering band in the OCT scan in Figs. 8(d) and 8(e), appearing in a depth up to $430\mu \mathrm{m}$. Again, the normal enamel (E) appears as light-dark changes in the right image part in Fig. 8(d).

Fig. 8

Upper row shows a single exemplary OCT B-scan (a) and generated 3-D scan (b) of the cervical area of tooth 24 of subject B (c). The OCT examination was performed contactless. Lower row presents the OCT results of the mineralization defect at tooth 24 of proband A (f) in 2-D (d,e). The second in vivo OCT measurement (d,e) was realized in contact mode by an additional distance piece (disposable item) at the distal endoscopic end. Abbreviations: E, enamel; D, dentin; G, gingiva. Scale bar in part (a), (d), and (e) corresponds to $300\mu \mathrm{m}$. Indicated dynamic range is valid for all depicted OCT images.

EOCT measurements were performed up to the region of the posterior teeth. A small fracture at the buccal side of tooth 16 in subject A is shown in Fig. 9. The B-scan and 3-D representation indicate that also the subjacent enamel layer is affected by the fracture. As already shown at the microcrack in Fig. 6(d), a shadowing effect due to higher backscattering of the incident light at the breakage can be found below the inset fractures.

Fig. 9

Exemplary B-scan (a) and 3-D representation (b) of a fracture at the buccal area of tooth 16 in subject A. The in vivo OCT imaging was executed in contact mode by the aid of a distal distance piece (disposable item). Scale bar in (a) corresponds to $300\mu \mathrm{m}$. Dynamic range of the depicted OCT cross-sectional scans amounts to 50 dB.

In addition, occlusal surfaces, such as the mesial-occlusal-distal composite restoration at tooth 45, in subject A, were imaged as exemplary presented in Fig. 10. The intersection between the restoration and the surrounding enamel shows a smooth transition without major cracks or gaps. The appearing bright section inside the restoration [cf. Fig. 10(b)] could be caused by an interlayer, possibly resulting from air entrapments or from varying scattering properties of adhesive and composite.

Fig. 10

(a,b) Representative B-scans of a mesial–distal–occlusal composite restoration at (c,d) tooth 45 in subject A. The in vivo OCT measurement was realized in contact mode by an additional distal distance piece (disposable item). Scale bar in (a) and (b) corresponds to $300\mu \mathrm{m}$. Dynamic range of the depicted OCT cross-sectional scans amounts to 50 dB.

3.2.

Imaging of the Oral Mucosa

For characterizing the performance of the handheld EOCT probe at the oral soft tissue, the mucosa at different parts of the oral cavity is imaged by means of a healthy volunteer (subject B in Sec. 3.1: female, 33 years, nonsmoker, occasional use of alcohol). Of particular interest for OCT imaging is the lining mucosa, which enables a larger depth of light penetration due to the absence of a corneous layer in comparison to the masticatory mucosa, such as the gingiva [cf. Fig. 6(e)], which is covered by keratinized stratified squamous epithelium. Figure 11 displays three exemplary B-scans of the buccal mucosa of subject B. With regard to the structural validation of oral mucosal tissue in OCT images by Hamdoon et al.,³⁸ the epithelial layer (EP) has a lower backscattering signal and appears more homogenous in comparison to the subsequent LP, which is confirmed by the imaging results in Fig. 11. The lower signal intensity of normal epithelium is related to the lower optical density and scattering properties.⁹ The LP corresponds to a dense fibrous layer with embedded blood vessels and nerves, which can be clearly differentiated from EP due to the higher backscattering signal and inhomogeneity in the OCT cross-sections. The EP and LP are divided by a basement membrane, which appears as a demarcation line between the two different backscattering signals of EP and underlying LP. OCT scans were acquired at the anterior [cf. Figs. 11(a) and 11(b)] and the posterior [cf. Fig. 11(c)] buccal mucosa. The epithelial thickness of the buccal mucosa is measured to be maximum $500\mu \mathrm{m}$ in the anterior part and of about $300\mu \mathrm{m}$ in the posterior region of the inside of the cheek.

Fig. 11

Single exemplary B-scans of the buccal oral mucosa in vivo, which are detected contactless at selected regions of the inner surface of the human cheek. (a, b) Buccal oral mucosa toward the human mouth in the anterior region of the oral cavity. (c) Buccal oral mucosa toward the oropharynx in the posterior region of the oral cavity. The OCT measurement was performed in noncontact mode. Abbreviations: EP, epithelium; LP, lamina propria. Scale bar: $300\mu \mathrm{m}$. Indicated dynamic range is valid for all depicted OCT images.

Additionally, the oral mucosa of the tongue and lateral floor of the mouth is imaged contactless by EOCT. Figure 12 demonstrates OCT cross-sectional images of the tongue body especially the inferior surface (a), the lateral border (b), and the curved upper surface (dorsum) (c). As a result, the EP thickness of the tongue strongly depends on its location and lowest for the inferior surface of the tongue with its smooth mucous membrane and thin stratified squamous epithelium [cf. Fig. 12(a)]. Additionally, the sublingual vessel network and the dense fibrous layer within the highly scattering LP are clearly visible in (a). The oral mucosa of the lateral border of the tongue in (b) shows a thicker EP, which is clearly differentiated from the subjacent LP with imaged minor and major blood vessels. In Fig. 12(c), the specialized oral mucosa of the surface of the tongue shows prominent epithelium ridges and tongue papillae. Due to the thick keratinized stratified squamous epithelium of the dorsal surface, the imaging depth is limited for which reason the LP is not visible in the OCT image [Fig. 12(c)]. An additional example for the lining mucosa is shown in Fig. 12(d), where the EP and the LP of the floor of the mouth are demonstrated.

Fig. 12

Single exemplary B-scans showing (a) the oral mucosa in vivo of the inferior surface of the human tongue (facies inferior linguae), (b) the margin of the tongue (margo linguae), (c) the dorsum of the tongue (dorsum linguae), and (d) the lateral floor of the mouth. All measurements are carried out in noncontact mode. Abbreviations: EP, epithelium; LP, lamina propria; B, blood vessel. Scale bar: $300\mu \mathrm{m}$. Indicated dynamic range is valid for all depicted OCT images.

Besides the representative OCT measurements of the buccal and lingual mucosa, the oral mucosa of the soft palate was imaged exemplarily with the developed OCT endoscope. For test purposes, the soft palate and the palatoglossal arch are examined in contact mode. In Fig. 13, the palatal mucosa is imaged by EOCT for three different regions of the soft palate. Figure 13(a) presents the oral mucosa of the anterior soft palate (toward the posterior border of hard palate), which obviously contains an LP with highly scattering dense connective tissue. This may be interpreted as part of the thin firm fibrous lamella named palatine aponeurosis. Below the dense fibrous layer, submucosal structures with integrated blood vessels are visible. In comparison, the oral mucosa detected at the palatoglossal arch is presented in Fig. 13(b) and shows a comparable EP, a thinner highly scattering fibrous layer as well as submucosal structures beneath. The oral mucosa of the transition part from the palatoglossal fold to the posterior buccal mucosa shows a varying EP thickness with a clearly determinable LP [cf. Figs. 13(c), 13(c1), and 13(c2)].

Fig. 13

(a–c) Cross-sectional images of oral mucosa at three selected regions of the soft palate, which are detected in contact mode. Oral mucosa of the anterior soft palate close to the posterior border of the hard palate (a) and corresponding overview image of the off-axis camera (a1). Oral mucosa of the central soft palate (b) and corresponding video guidance image (b1). Oral mucosa of the palatoglossal arch close to the posterior buccal mucosa in 2-D (c) and 3-D (c1,c2). Abbreviations: EP, epithelium; LP, lamina propria; B, blood vessel. Scale bar: $300\mu \mathrm{m}$. Indicated dynamic range is valid for all depicted OCT images.

In addition, the oral mucosa of the palatoglossal arch was imaged via contactless EOCT. The results are presented in Fig. 14, where the EP and subsequent LP are obviously differentiable.

Fig. 14

Contactless OCT detection of oral mucosa of the palatoglossal arch of the soft palate with two capillary branches. (a) 2-D OCT with both capillary branches (CN1, CN2) embedded in the papillary layer of the LP. (b,c) OCT image series of the two capillary branches showing cross-section in front, through, and after the small capillary network. Abbreviations: EP, epithelium; LP, lamina propria; CN, capillary network. Scale bar: $300\mu \mathrm{m}$. Indicated dynamic range is valid for all depicted OCT images.

Moreover, two capillary branches (CN1 and CN2) lying relatively close to each other are detected within the papillary layer of the LP. Both capillary branches running to the papilla are detected separately in 3-D. Representative B-scans of the OCT volume scans are presented as image series in Figs. 14(b) and 14(c), showing the respective cross-section in front, through, and after the capillary networks CN1 and CN2 within the superficial layer of the LP.