2.1.

Tooth Samples and Chemicals

Human premolars (n=15) were collected from orthodontic patients at the Dental Clinic of the Faculty of Dentistry, University of Manitoba. Approvals from the human ethics committees of the Institute for Biodiagnostics (National Research Council Canada), University of Manitoba and Dalhousie University were obtained prior to sample collection. All teeth were examined in patients before extraction and no discoloration of the marginal ridge was observed. Remaining soft tissue on extracted teeth was removed by scaling and the samples were thoroughly rinsed with water. Teeth were preserved in sterile filtered de-ionized water until measurement. Each tooth sample was radiographed and independently re-assessed ex vivo by two clinical investigators at the University of Manitoba and Dalhousie University, respectively. Control caries-free teeth had no visible decalcification or demineralization. Incipient carious teeth included regions of decalcification with intact surfaces and opacity of enamel. Among the 15 extracted teeth, 10 were identified to be carious teeth with at least 15 caries sites in total. The teeth were used for spectroscopic measurements without any treatment. Synthetic hydroxyapatite in powder form was obtained from Sigma-Aldrich Inc. (St. Louis, MO, USA) and used for confirmation of Raman spectral peak assignments of hydroxyapatite within tooth enamel.

2.2.

Optical Coherence Tomography

Optical coherence tomography image slices were acquired on an OCT-2000 system with software Revision A (Humphrey Systems, Dublin, CA, USA) and equipped with a superluminescent diode source. The source has a central wavelength of 850 nm and the laser spot size is 10–20 μm. The optical power at the sample was 750 μW for all image sets. Both the axial and transverse resolutions were 10–20 μm. The OCT-2000 system has an x–y galvanometer pair for beam steering and an integrated camera for sample viewing during data collection. For image acquisition, the galvanometer pair was limited to straight-line collection only. The shortest lateral scan length (∼2.0 mm) of the system was used to provide images of high pixel density. The scan rate was 100 A scans/s.

For these ex vivo studies, tooth samples were imaged in an upright position by affixing the apical root portion of the tooth to the microscope slide using dental rope wax. The samples were imaged using a free-space coupled arrangement. The laser line was focused to the thinnest line on the sample image and during acquisition, the polarization was matched to the sample in order to optimize the OCT signal strength. OCT data were exported into MATLAB (The Mathworks Inc., Natick, MA, USA) and interpolated using a bilinear filter method within MATLAB to generate false-color images. The scan length was corrected for nonophthalmologic uses of the system by imaging a linear scale. The depth distance was corrected by dividing the depth obtained from the system (assuming imaging in air) by a value of 1.6, the refractive index of tooth enamel.⁴⁰

2.3.

Raman Microspectroscopy and Fiber-Optic Raman Spectroscopy

Raman spectra were acquired on a LabRamHR confocal Raman microspectrometer (Horiba Jobin Yvon, Edison, NJ, USA) operating with near-infrared laser excitation at 830 nm (Lynx series TEC 100 diode laser, Sacher Lasertechnik GmbH, Marburg, Germany), (Fig. 1). In brief, the Raman microspectrometer consists of an Olympus BX41 microscope equipped with a motorized XYZ stage, a spectrograph with 300 lines/mm grating and an air-cooled CCD detector optimized for the NIR region. Spectra were also acquired using a fiber-optic Raman probe (830 nm excitation, InPhotonics, MA, USA) that was interfaced to the LabRamHR spectrometer. Laser powers at the sample were 24, 39, 54, 48, and 52 mW under ×100 (Olympus LMPlan IR), ×50 (Olympus MPlan), ×10 (Nikon), and ×5 (Leica HC PL Fluotar) microscope objectives and fiber-optic probe, respectively. For microspectroscopy, the confocal hole size was set at 800 μm and the slit size at 100 μm. The spectral resolution was 4 cm⁻¹ for the spectra acquired with the microscope objectives and 7 cm⁻¹ for the fiber-optic probe. These values were determined using an 841 nm neon line. A polarization scrambler is placed in the Raman collection path in order to eliminate potential artifacts from polarization-sensitive components (e.g., diffraction grating) (Fig. 1). LABSPEC (version 4.12, Horiba Jobin Yvon, Edison, NJ, USA) software accompanying the LabRamHR system was used for spectrometer control and data acquisition. For Raman microspectroscopy, the tooth surface not being studied was placed lying on a microscope slide and then secured with dental wax. Wax was applied only at the apical root to avoid wax contamination at the surfaces of interest. Tooth surfaces to be examined were positioned approximately normal to the laser beam (Fig. 1). For fiber-optic sampling, tooth samples were positioned upright on a microscope slide and the fiber-optic probe was placed at a position 5 mm away (optimal laser focus) from the tooth surface. The sampling position was optimized by a XYZ translator assembled in-house. A camera coupled with an ×10 objective lens was used for monitoring the positioning and for capturing photomicrographs of the sampled locations.

Figure 1

Schematic diagram of the Raman microspectroscopic system illustrating the laser excitation (solid line) and Raman signal detection (dashed line) paths for acquiring tooth spectra in a 180° geometry. Abbreviations are provided in the inset for each optical element. The linear polarizer (P) is only used when performing polarized Raman experiments.

For our studies, the orientation of the enamel crystals within tooth samples were defined with respect to the laser beam as has been described by Tsuda and Arends.³⁶ Raman spectra obtained from the cross-sectioned surface of longitudinally sectioned (i.e., crown to root direction) tooth samples were defined as acquired in a transverse excitation/detection mode [Fig. 2(A)]. Raman spectra recorded from unsectioned whole tooth were defined as acquired in a normal excitation/detection mode [Fig. 2(B)]. Spectra were measured using 30 s acquisition time with 15 accumulations for the ×5 objective or 15 s acquisition time with 15 accumulations for the other objectives and the fiber-optic probe in order to generate spectra of good signal to noise ratios. For point mapping experiments using the ×10 objective, the laser spot size was about 37×25 μm². Spectra were acquired at 140 μm steps along the x axis and 113 μm steps along the y axis resulting in a 10×10 array map covering an area of 1260×1017 μm².

Figure 2

Diagrammatic representation of the two spectral acquisition configurations using (A) transverse excitation/detection on a longitudinally sectioned (crown to root direction) tooth, and (B) normal excitation/detection on an unsectioned whole tooth. Drawing illustrates the actual sample orientation under the microscope objective.

Polarized Raman microspectroscopic measurements of tooth samples were acquired by placing a NIR (780–1250 nm range) linear polarizer (Melles Griot, Irvine, CA, USA) in the laser excitation path after the radiation exits the laser head (Fig. 1) to obtain linearly polarized light from the originally scrambled laser polarization (Lynx series TEC 100, Sacher Lasertechnik GmbH, Marburg, Germany). One set of measurements was acquired with the linear polarizer in one orientation and a subsequent set of measurements was acquired with the linear polarizer rotated 90° such that the second polarization direction is orthogonal to the first polarization direction. All other measurement conditions (e.g., optics, sample orientation, etc.) were left unchanged between the two measurement sets. Similar to the other nonpolarized measurements, a polarization scrambler is placed in the Raman collection path in order to eliminate potential artifacts from polarization-sensitive components.

2.4.

Raman Spectroscopic Data Analysis

Background Raman spectrum, acquired with no sample in place and with all other experimental conditions unchanged, was subtracted from sample spectra to correct for background signal arising from optical elements in the laser path. The spectrum of a luminescent green glass reference, calibrated previously with a NIST traceable reference tungsten halogen lamp (The Eppley Laboratory, Inc., Newport, RI, USA) of known temperature, was used to correct the sample spectra for the instrument response function.^{41 42 43} The majority of the Raman spectra (sound enamel and carious enamel) acquired contained a minor sloping baseline as the background [Fig. 3(A)]. A few spectra of carious enamel contained a large background fluorescence [Fig. 3(B)]. For semiquantitative spectral analyses and mapping experiments, the spectra were first baseline corrected using a sixth order polynomial fit through the spectra at 375, 512, 700, 840, 1200, and 1400 cm⁻¹. Figure 3 illustrates representative spectra before and after the backgrounds were removed by subtracting the fit of the sixth order polynomial. The corrected spectra were then normalized to the 959 cm⁻¹ peak. Peak ratios were determined by taking the intensities of the various Raman bands of interest. In a similar approach, an intensity ratio map was generated from the mapping data. The resulting Raman map presented was smoothed using the bilinear interpolation method for two-dimensional data in MATLAB.

Figure 3

Representative Raman spectra illustrating the background fluorescence acquired from (A) the majority of sound and carious dental enamel and (B) from a few number of carious sites. For both panels, the upper solid traces illustrate the original spectrum. The dashed traces are the sixth order polynomial fits that were used for correction and the lower solid traces are the results after subtracting the polynomial fits of the background fluorescence.

3.1.

Optical Coherence Tomography of Dental Caries

Figure 4 shows representative false-color OCT images obtained from a sound tooth and a tooth surface containing two clinically confirmed incipient lesions also known as white spots. The images present the lateral scan position versus the imaging depth with higher intensity correlating with greater light backscattering. In the image of the sound tooth [Fig. 4(A)], an intense light backscattering is observed at the tooth surface. This represents the scattering of the light due to the change of refractive index as the light transitions from air to the tooth enamel surface. For the sound surface, beyond the initial first few microns, the light backscattering rapidly decays with no further changes in intensity deeper into the enamel. This image suggests that the surface is intact with no structural defects, increased porosity or loss of mineral structure. There is no evidence of a further scattering boundary deeper into the tooth that would suggest the presence of the dentin–enamel junction (DEJ). This feature, however, has been observed by other groups performing OCT imaging of tooth.^{14 16} The absence of an observable scattering boundary representing the DEJ on our images is possibly due to the large enamel thickness observed at the region sampled and the attenuation of the scattered signal as it penetrates the enamel; enamel is known to weakly scatter near-infrared light (850 nm in our studies).⁴⁰ Other OCT images acquired of the enamel closer to the cemento–enamel junction (i.e., borderline between the crown and root complex) have demonstrated the DEJ (figure not shown for brevity). Therefore, it is likely the enamel thickness that limits the observation of the DEJ in our OCT images. In contrast to the image of the sound enamel surface, the image [Fig. 4(B)], taken of the carious sites portrays diffuse scattering in two triangular-shaped zones immediately below the surface areas. Once again there is intense light backscattering at the tooth surface indicating that the incipient lesions have intact surfaces. The diffuse scattering intensity in the region below the surface is due to the occurrence of multiple scattering and indicative of an area of higher porosity within an otherwise dense enamel structure. This suggests that demineralization has occurred below the intact surface as occurs with early dental caries formation. Similar imaging results were reported by Fried etal.¹⁹ using PS-OCT on natural interproximal lesions. Based on the OCT image, it is estimated that the deepest area of the lesions is approximately 290 μm deep. The triangular-shaped region below the surface and this depth estimation is consistent with histological studies in the literature that have shown a similar triangular-shaped lesion body with 100–250 μm depth typical of “white spot” carious lesions.^{11 12 13} OCT is able to provide morphological information of near-surface tissue structures and defects and is particularly sensitive to changes in refractive index as the light interacts with the sample. It is therefore a good first approach for examining tooth samples to screen for early dental caries and estimating the lesion depth. From a clinical perspective, lesion depth is useful for determining the extent of caries activity and in aiding the decision to surgically restore or promote remineralization. In addition, the depth can be used to monitor remineralization and to evaluate the arrest of further caries progression.

Figure 4

Representative false-color OCT images of lateral scan length as a function of depth from (A) sound enamel tooth surface (minimum light back scattering with depth) and (B) carious enamel surface containing two lesions (significant light backscattering). The carious lesion sites are indicated with the arrows.

3.2.

Raman Microspectroscopy of Sound and Carious Enamel

Following the identification of a possible carious lesion by OCT, false-positive results can be reduced by confirming the presence of caries using Raman spectroscopic characterization. Figure 5 illustrates representative Raman spectra, obtained with an ×10 objective, of sound enamel and carious enamel, both from unsectioned whole teeth, and of synthetic hydroxyapatite (OHAp). The major band positions and tentative assignments of enamel and OHAp spectra are summarized in Table 1 and are compared with those previously reported. Our Raman spectra show good agreement with the literature values in terms of the respective Raman shifts. An examination of the Raman spectra from sound and carious enamel revealed differences in the relative band intensities of various Raman peaks, however, no new bands, band shifts, or disappearance of bands were evident. The symmetric stretching vibration (ν₁ mode) of phosphate (PO ₄ ³⁻) at 959 cm⁻¹ dominates both sound and carious enamel spectra. The peak position is characteristic of carbonated biological apatite found in bone, dentin and enamel.^{21 24 29} A similar peak in the spectrum of synthetic hydroxyapatite has a peak maximum at 962 cm⁻¹. The higher Raman shift of this peak indicates a higher crystallinity of the synthetic hydroxyapatite as compared to carbonated hydroxyapatite. In carbonated hydroxyapatite, CO ₃ ²⁻ has been known to substitute PO ₄ ³⁻ yielding type-B carbonated hydroxyapatite; in type-A CO ₃ ²⁻ substitution, the OH ⁻ is replaced in the hydroxyapatite crystal.^{44 45} The incorporation of CO ₃ ²⁻ into the hydroxyapatite crystal deforms the crystal structure causing a decrease in crystallinity. There was no obvious change in the full-width at half-maximum (FWHM) of the 959 cm⁻¹ peak between sound and carious enamel spectra. The FWHM is ∼10.0±0.24 cm⁻¹ and corresponds to earlier studies that have examined the FWHM of enamel.⁴⁶ Studies involving infrared absorption and Raman studies have reported the broadening of this peak upon incorporation of CO ₃ ²⁻ (type-B) into synthetic hydroxyapatite.^{47 48} Since an early event of demineralization is the dissolution of CO ₃ ²⁻ by acid attack,⁴⁹ one might expect the spectra of carious lesions to have an increased ν₁ PO ₄ ³⁻ Raman shift and decreased peak width compared to spectra of sound enamel. Both of these features would be indicative of increased hydroxyapatite crystallinity. Such changes were not observed. Our studies involve biological hydroxyapatite, however, the crystallinity differences reported in the literature were from studies involving synthetic hydroxyapatite crystals.

Figure 5

Representative microspectroscopic Raman spectra of (a) sound human enamel, (b) carious human tooth enamel and (c) synthetic hydroxyapatite (OHAp) acquired with ×10 objective. Spectra are offset for clarity. The asterisks indicate peak positions. See discussion for further details.

Table 1

An examination of the 1069 cm⁻¹ peak assigned to both PO ₄ ³⁻ and type-B CO ₃ ²⁻ of apatite also show no obvious changes between sound and carious enamel despite the loss of carbonate with demineralization. Chemical analyses of different histological zones of the enamel lesion showed a carbonate level of 1–2% (wt/wt) in the lesion body compared to an average of 2% (wt/wt) in the outside layers of sound enamel.^{50 51} It is not surprising that Raman spectroscopy is unable to detect the difference at this low concentration level. Spectral analyses is further complicated by the fact that the peak at 1069 cm⁻¹ is a combination of PO ₄ ³⁻ and type-B CO ₃ ²⁻ vibrations.

In spite of these findings, changes are visibly observed with the formation of white spot lesions. There should then be some underlying biochemical or structural alterations giving rise to the clinically detected differences. The various other phosphate peaks were therefore examined more closely. In particular, peak intensity ratios were determined to be a better method for interpreting such spectroscopic data. This approach was chosen since Raman spectroscopy is known for its sensitivity toward sampling geometry such as laser focusing distance, sampling angles and the topography of the sample surface. This is especially true for (confocal) Raman microspectroscopy where the sampling volume is very small and any minor variations in sampling condition can alter the collection efficiency, thus affecting the signal intensity.

Figure 6 shows representative Raman spectra of sound and carious enamel in the regions of 350–700 cm⁻¹ [Fig. 6(A)] and 800–1200 cm⁻¹ [Fig. 6(B)]. Symmetric bending vibrations (ν₂ mode) of PO ₄ ³⁻ give rise to two major peaks at 431 and 446 cm⁻¹. In the sound enamel spectrum, the 446 cm⁻¹ peak has a greater intensity than the 431 cm⁻¹ peak. These two bands are also present in the caries spectra, however, their relative intensities are reversed with the 431 cm⁻¹ peak intensity greater than the 446 cm⁻¹ peak. A series of bands at 579, 590, 608, and 614 cm⁻¹ (shoulder band) are assigned to asymmetric bending vibrations (ν₄ mode) of PO ₄ ³⁻. The band pair at 579 and 590 cm⁻¹ shows an intensity profile change similar to that observed for the ν₂ mode with the 590 cm⁻¹ band of greater intensity in the caries spectrum but weaker intensity in the sound enamel spectrum. The two bands at 608 and 614 cm⁻¹ do not demonstrate obvious changes between sound and carious enamel spectra. The asymmetric stretching vibration (ν₃ mode) of PO ₄ ³⁻ constitutes a more complicated region of the spectra [Fig. 6(B)]. Raman bands were observed at 1023, 1043, 1052, and 1069 cm⁻¹. The intensity of the 1043 cm⁻¹ band increases noticeably in the caries spectrum whereas the other three bands do not show prominent intensity variations. The numerous tooth samples examined showed various degrees of spectral change between sound and carious sites, however, the bands at 431, 590, and 1043 cm⁻¹ showed consistent characteristic differences. These changes are highlighted in Fig. 7(B) where the intensity ratios of various bands relative to the 959 cm⁻¹ band are plotted against sampling locations indicated in the photomicrograph of Fig. 7(A). This figure depicts an enamel surface containing a carious lesion. The carious lesion appears as a white area on the image with the marked dots representing various sampling locations. The intensity ratio plot shows enhanced 431, 590, and 1043 cm⁻¹ bands relative to the 959 cm⁻¹ band at the carious lesion whereas the other bands do not show significant intensity changes. The same finding is observed in the Raman intensity ratio image map based upon the 1043 cm⁻¹/959 cm⁻¹ peaks obtained from a 10×10 point mapping study [Fig. 7(C)]. Both the size and the location of the carious lesion on the Raman map match very well with the photomicrograph. This result further confirms the validity of the Raman bands selected for caries detection. Overall, the findings of Fig. 7 describe a method of contrasting sound and carious enamel based upon Raman signal intensity ratios of various PO ₄ ³⁻ vibrations. Hill and Petrou³⁹ have previously reported that carious lesions can be distinguished from sound enamel based on their Raman spectra. However their method was not fully based on the enamel’s intrinsic Raman signal but rather on the level of the luminescence background upon illumination with 785 nm laser excitation. For the most part, our studies using laser excitation further to the near-infrared region (830 nm) did not show any significant background luminescence. A similar reduction of background fluorescence was also apparent in spectra acquired by Hill and Petrou in an earlier study using 1064 nm laser excitation.³⁸ Therein, the study focused on the 960 cm⁻¹ peak as well as the 930 and 1900 cm⁻¹ peak intensities for evaluation of the luminescence background.

Figure 6

Representative microspectroscopic Raman spectra in the region of (A) 350–700 cm⁻¹ and (B) 800–1200 cm⁻¹ of sound enamel (a and c) and carious enamel (b and d). Spectra were acquired with ×10 objective and offset for clarity. The asterisks indicate peak positions.

Figure 7

(A) Photomicrograph of a human enamel tooth surface containing a carious lesion (white area). The shaded dots represent the sampling locations of the spectra acquired across the surface and used to generate (B) the peak intensity ratio plot of various phosphate peaks relative to the 959 cm⁻¹ peak. (C) Raman 2-D image map of the same lesion area illustrated in (A). The smoothed Raman spectral imaging map was generated using the intensity ratios of the 1043 cm⁻¹/959 cm⁻¹ peaks and plotted as a function of the 10×10 point mapping array. (See Sec. 2 for details.)

In order to determine a rationale for the characteristic intensity changes observed at 431, 590, and 1043 cm⁻¹ between sound and carious enamel, we compared our spectra more closely with those reported in the literature. It was noted that our Raman spectra of sound enamel from whole unsectioned teeth did not have the same relative peak intensities as reported in the literature of sound enamel from cross-sectioned tooth samples. The difference between these studies lies in the sampling configuration in which Raman spectra were acquired. Representative Raman microspectroscopic data of sound tooth enamel measured using two different sampling geometries are shown in Fig. 8. The upper spectrum was acquired on a longitudinally sectioned tooth sample and known as a transverse excitation/detection (exc./det.) mode whereas the lower spectrum was acquired on an unsectioned whole tooth termed the normal exc./det. mode. The transverse and the normal sampling arrangements are demonstrated in Fig. 2 and in the inset in Fig. 8. These two spectra show differences in several band intensities, which are highlighted with asterisks. The transverse spectrum differs from the normal spectrum with relatively higher Raman signal intensities at 431, 590, and 1043 cm⁻¹. These same intensity differences were also observed between sound enamel spectra and carious enamel spectra when measured using a normal exc./det. configuration. The spectrum of carious enamel measured in the normal exc./det. configuration is an intermediate between the spectra of sound enamel measured with the two sampling geometries, transverse versus normal. Considering that the biochemical composition does not change when the sampling arrangement has changed and given that the Raman signal is sensitive to both the (bio)chemical composition and structure of the sample, it is believed that the spectral differences between the two measurement configurations and between sound and carious enamel arise from structural characteristics of the enamel rods.^{36 37} Tooth histology indicates that hydroxyapatite crystals are bundled into long rods with ∼4–7 μm diameter.^{49 52} A small portion of the enamel rods are randomly oriented but the majority are highly oriented within the enamel layer with their c axis (the longer axis) approximately perpendicular to the natural tooth surface.^{36 49} This orientation is observed on the entire crown surface, i.e., at the top (occlusal or incisal surface) and at the sides (proximal surfaces) of the tooth crown.⁴⁹ Previous polarized Raman studies have indicated that changes in the spectral profile of enamel crystals reflect changes in the c-axis orientation.^{36 37} Therefore, we propose that the spectral differences observed between sound and carious enamel are due to the hydroxyapatite crystallite orientation in the enamel. The demineralization process of caries formation results in changes of the enamel crystallite morphology or a loss of preferred enamel crystallite orientation which is reflected in the Raman spectrum. Polarized Raman spectroscopy has been shown to be useful in determining the enamel crystallite orientation and the structural symmetry.^{36 37} Preliminary polarized Raman studies on sound and carious enamel were performed to explore the hypothesis that demineralization results in alterations in enamel crystal orientation. Figure 9 shows representative polarized Raman spectra of sound enamel (traces a & b) and carious enamel (traces c & d) acquired with the linear polarizer at one polarization orientation (traces a & c) compared to the orthogonal polarization orientation (traces b & d). It is observed that the intensity of the phosphate ν₂ (431 cm⁻¹), ν₄ (590 cm⁻¹), and ν₃ (1043 cm⁻¹) peaks of sound enamel alters with the change in the polarization direction of the laser excitation. These peaks are therefore sensitive to the change in laser polarization suggesting that sound enamel is optically anisotropic. In contrast, these same peaks in Raman spectra of carious lesions do not change significantly when the laser polarization direction is rotated. Therefore the peaks are much less sensitive to the changes in laser polarization and this indicates that carious enamel is optically isotropic due to scrambling of the hydroxyapatite crystallite orientation upon demineralization. These observations support the suggestion that crystallite structural orientation changes are responsible for the spectral changes observed between sound and carious enamel.

Figure 8

Representative microspectroscopic Raman spectra of human tooth enamel using (a) transverse exc./det. configuration and (b) normal exc./det. sampling mode configuration. Measurements were acquired with a ×10 objective. Peaks highlighted with asterisks show major intensity differences between the two sampling modes. Inset: schematic illustrations of transverse (sectioned tooth) and normal (whole tooth) exc./det. sampling arrangements. Laser beam is 90° to the sample’s surface in both cases.

Figure 9

Representative polarized Raman spectra of sound enamel (a and b) and carious enamel (c and d) acquired with the linear polarizer in one polarization direction (a and c) and the orthogonal polarization orientation (b and d). The asterisks highlight peak positions.

3.3.

Fiber-Optic Raman Spectroscopy of Sound and Carious Enamel

So far, we have demonstrated that carious enamel can be distinguished from sound enamel at the microspectroscopic level using the intensity changes of the hydroxyapatite Raman bands. We proceeded to investigate the suitability of a fiber-optic Raman probe for caries detection. Representative fiber-optic Raman spectra (not corrected for background fluorescence) of carious and sound tooth are illustrated in Fig. 10 with band assignments listed in Table 1. The major bands identified in spectra acquired with the fiber-optic probe correspond well to those observed from microspectroscopy. The overall spectral pattern of the fiber probe spectra resembles that of the microspectroscopic spectra. However, a slightly higher level of background luminescence is observed with the spectra possessing lower spectral resolution (∼7 cm⁻¹) as shown by the unresolved peak at 1053 cm⁻¹. Comparing sound enamel and carious enamel spectra, similar increases in Raman peak intensities at 433, 590, and 1043 cm⁻¹ are detected using a fiber-optic probe as with Raman microspectroscopy. The degree of changes is however less than that detected with the microscope objective and is likely the result of a larger sampling depth obtained by the fiber probe. The larger sampling depth obtained with the fiber probe indicates that the region beyond the caries lesion and into the healthy enamel layer was possibly measured in the Raman spectra. As such, any spectral contribution from the carious lesion is gradually diminished in the overall spectrum in turn resulting in slightly decreased degree of spectral discrimination between sound and carious enamel.

Figure 10

Representative fiber-optic Raman spectra (not corrected for luminescence background) of (a) carious enamel and (b) sound enamel acquired with an InPhotonics Raman probe for 830 nm laser excitation. Spectra were normalized to the 959 cm⁻¹ peak and were offset for clarity.

The greater sampling depth of the fiber-optic probe is also supported by the observation of a more intense band at 1070 cm⁻¹ for fiber probe spectra compared to spectra acquired by microscopy. As mentioned earlier, this peak arises from PO ₄ ³⁻ and type-B CO ₃ ²⁻ of hydroxyapatite. Carbonate content is known to increase within the tooth progressing from the enamel surface to the DEJ where it reaches its maximum.^{29 49 53} In Fig. 11, we present Raman microspectroscopy spectra measured of sound whole tooth enamel with a series of increasing magnification microscope objective (×5,×10,×50,×100) with increasing numerical apertures and therefore decreasing sampling volumes. Focusing on the 750–1200 cm⁻¹ region, we observe that the CO ₃ ²⁻ intensity at 1069 cm⁻¹ is indeed decreasing continuously from the ×5 spectrum to the ×100 spectrum. The results with a low magnification objective and the fiber-optic data are in agreement with the known rise in carbonate content deeper into the tooth crown.

Figure 11

Representative Raman microspectroscopic spectra in the 750–1200 cm⁻¹ region of sound enamel measured with (a) ×5, (b) ×10, (c) ×50 and (d) ×100 microscope objectives.

The larger sampling depth of the Raman fiber probe resulted in a slight decrease in the spectral differences between sound and carious enamel. For this reason then, it is desirable to use a fiber probe with pseudoconfocal arrangement where the measurement volume more closely approximates that of the ×10 microscope objective. With such sampling depths, healthy and early decaying enamel can readily be distinguished. Although useful for characterizing and confirming demineralization, using such a Raman probe alone would not be practical for screening in order to locate possible carious sites. The small sampling of Raman point spectroscopy would result in a greater chance of missing the lesion. For this reason, it is beneficial to couple OCT imaging with Raman spectroscopy. For the purposes of detecting suspicious carious sites, OCT would be useful for rapidly screening larger areas than that interrogated with the Raman probe. The region covered by the current OCT system is approximately 1 mm across. Although a small range, this is sufficient to probe the approximately 2 mm×2 mm area found below (gingival to) the contact points of adjacent teeth where such interproximal caries develop. OCT can therefore be used to sweep across this region to rapidly identify possible early carious sites as well as to determine the lesion depth. Based upon OCT morphological guidance, Raman spectroscopy can then be used to examine the suspected lesion to provide biochemical information. For example, a Raman probe can be used to sample various points along the OCT imaging line. Then based upon the ratios of the various Raman peaks, it can be determined rapidly whether the region exhibits spectral properties indicative of (early) carious enamel and distinguished from sound enamel. The Raman data therefore confirms the scattering aberration detected by OCT to be a carious lesion. The combination of both techniques overcomes the limitation of using each technique alone leading to results of better sensitivity and specificity.

In order to examine the reproducibility of fiber-optic Raman spectroscopy for caries characterization, multiple-point measurements were acquired along a line on a tooth surface containing two carious lesions. The same lesion was previously examined by OCT and shown in Fig. 4(B). Figure 12(A) highlights the points sampled (shaded dots shown on the image in a linear fashion) and Fig. 12(B) shows the respective intensity ratio of various peaks relative to the ν₁ PO ₄ ³⁻ peak at each sampling location. The increased intensity ratio correlates well with the locations of the carious lesions, as previously observed with the microspectroscopic study [see Fig. 7(B)]. These findings also correspond well with the OCT imaging where regions of carious lesions demonstrated increased light backscattering intensity. Therefore, the spectroscopic contrast between the carious and sound enamel present in the microscopic spectra is largely preserved in the fiber-optic measurements. This observation supports the possibility of transferring this technology from a lab bench study to a clinical application.

Figure 12

(A) Photomicrograph of a tooth surface containing two carious lesions (white areas). Shaded dots represent actual Raman sampling locations with a fiber-optic Raman probe. (B) Intensity ratio plot of various Raman peaks (431, 589, 1042, and 1070 cm⁻¹) relative to the 961 cm⁻¹ peak corresponding to the sampling locations shown in (A).