1. Introduction

In spite of the introduction of fluoride in drinking water, use of fluoride toothpaste, fluoride treatment in dental clinics and increased emphasis on effective home care strategies for the patient, dental caries (i.e., dental cavities or dental decay) remains the most common oral disease that people in the developed countries have experienced. The patterns of caries development have changed over the years, to those with smaller lesion sizes and slower progression rates. Further, new carious lesions are commonly found just below the interproximal contact sites (e.g. between adjacent teeth), an area that is difficult to examine. These characteristics of the caries problem magnify the limitations of caries detection through conventional diagnostic methods that involve subjective clinical criteria (colour, “softness”, resistance to removal) and the use of diagnostic tools such as the dental explorer and dental radiographs. These conventional methods are adequate for detection of larger, possibly cavitated lesions, but due to poor specificity and sensitivity, they are not suitable for detection of early stage, non-cavitated lesions. In addition, these clinical methods do not possess the ability to address the dynamic nature of the demineralization-remineralization process during caries formation [1]. Therefore, better diagnostic tools are needed to detect early non-cavitated carious lesions and to monitor their activity (changes in severity). Over the past few decades, much effort has been dedicated to the development of new caries detection tools based on optical methods with improved sensitivity and specificity [2,3]. Among those include direct digital radiography (DDR), digital imaging fibre-optic trans-illumination (DIFOTI), quantitative light-induced fluorescence (QLF), laser-induced fluorescence [2], multi-photon imaging [4], infrared thermography [5], terahertz imaging [6], optical coherence tomography (OCT) [7] and Raman spectroscopy [8,9]. Some of these techniques have also led to commercial products, such as DIFOTI from Electro-Optical Sciences, Inc., QLF^TM from Inspecktor Research Systems and DIAGNOdent^TM from KaVo. Just recently, a new imaging technique similar to the visible light trans-illumination technique but using near-infrared (NIR) light, was reported to be useful for detection of interproximal and occlusal caries [10,11]. The 1310 nm NIR light used in the study was shown to be better than visible light used in DIFOTI due to higher light transmission thereby providing higher contrast between carious and sound enamel. The capability for quantitating caries severity remains a problem for this technique. Another potentially useful optical technique for caries detection is optical coherence tomography (OCT) and it has received much attention recently due to its capability for realtime depth imaging of internal tooth structure at high spatial resolution. The back-scattered light intensity provides near surface morphology, therefore providing relevant information for incipient (i.e. early non-cavitated) caries detection. Lately, polarization-sensitive OCT (PS-OCT) has been shown to provide added contrast between sound and carious enamel [12]. However, using OCT imaging alone could lead to false-positive results and the technique is also subject to inter- and intra-examiner variability.

In our previous study, we reported a new multi-modal approach based on optical coherence tomography (OCT) and Raman spectroscopy for detecting and characterizing early dental caries [8]. While OCT focused on detecting morphological changes of the tooth surface structure during tooth demineralization, Raman spectroscopy provided biochemical characterization of hydroxyapatite (Ca₁₀(PO₄)₆(OH)₂), the major mineral component of tooth enamel. Several Raman peaks of PO₄ ^3- vibrations arising from hydroxyapatite showed an increased peak intensity in the spectra of carious lesions when compared with sound enamel spectra, particularly at 431 cm^-1, 590 cm^-1 and 1043 cm^-1. Based on information derived from non-polarized Raman spectra, it was proposed that such spectral variation resulted from changes in morphology and/or orientation of enamel rods within the carious lesions. Preliminary polarized Raman data obtained in that same study also showed that the Raman spectra of carious lesions were less sensitive than those of sound enamel to the rotation of laser polarization direction. Therefore it was believed that a more thorough study using polarized Raman spectroscopy would provide insight relevant to the understanding of the caries process. This information would support an improved detection methodology for early caries using Raman spectroscopy.

Polarized Raman spectroscopy is known as a useful technique for studying molecular structure. For randomly oriented molecules in solution, the depolarization ratio (ρ) is mainly dependent on vibrational symmetry and therefore can aid in peaks assignments. In solid samples such as crystals, when molecular orientation is known relative to the polarization of the laser beam’s electric field, the depolarization ratio (ρ) is strongly influenced by molecular alignment, therefore, ρ can provide additional structural information. Recent applications of polarized Raman spectroscopy on solid samples include determination of molecular orientation of uniaxially oriented polymers [13], protein structures [14] and carbon nanotubes [15]. Polarized Raman spectroscopy was also employed by Tsuda et al [16] and by LeRoy et al [17] to study the orientation of tooth enamel rods, however, no study on early dental caries based on polarized Raman spectroscopy has been reported in the literature. Since the majority of enamel rods have preferred orientation within the tooth, any orientational changes and/or scrambling of enamel rods caused by caries activity will likely alter the sample’s polarized Raman spectral profile. This report describes a polarized Raman spectroscopic methodology for discriminating between carious and sound enamel.

2. Materials and methods

2.1 Tooth samples

Thirteen extracted human premolars were acquired at the Oral Surgery Clinic, Faculty of Dentistry, University of Manitoba. These teeth were extracted from consenting patients who were undergoing extractions for orthodontic reasons. Prior to teeth collection, approvals for using human tissue samples were obtained from the human ethics committees of all participating institutes involved in this study. The teeth were examined by one dental clinical investigator for signs of early caries at the proximal sites prior to extraction. Remaining soft tissue on extracted teeth was removed by scaling and the samples were thoroughly rinsed with water. Teeth were then preserved in sterile filtered de-ionized water until measurement. Each tooth sample was radiographed and independently re-assessed ex vivo by two dental clinical investigators at the University of Manitoba and Dalhousie University, respectively. Caries-free teeth had no visible decalcification or demineralization while incipient carious teeth included regions of decalcification with intact surfaces and opacity of enamel (white spots when teeth surfaces were dry). More than twenty carious lesions were identified in 13 teeth. These samples were used for spectroscopic measurements without further treatment.

2.2 Polarized Raman microspectroscopy and polarized Raman spectral imaging

The Raman instrumentation and the tooth sampling geometry used in this study have been described previously [8], with the addition of a NIR (780–1250 nm range) linear polarizer (Edmund Optics, Blackwood, NJ, USA) at the exit of the laser radiation and a NIR polarization analyzer placed after the notch filter—see Fig. 1. The linear polarizer and the polarization analyzer were located in this manner so that the initial laser radiation and the detected Raman scattered photons could be independently polarized. Unsectioned tooth samples were placed lying on a microscope slide with the surfaces to be studied positioned approximately normal (i.e., 90°) to the laser beam in a backscattering sampling geometry as shown in Fig. 1. The polarization-dependent grating efficiency was corrected by inserting a polarization scrambler before the entrance slit of the spectrograph. In brief, 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 TEC-100 or Tiger TEC-500, Sacher Lasertechnik GmbH, Marburg, Germany). Both the Lynx and the Tiger lasers are external cavity diode lasers with similar laser beam profiles and can provide 100 mW and 500 mW power at the laser head at 830 nm, respectively. With the NIR polarizer in place, laser power at the sample was measured to be 12 mW and 40 mW under a 10x (Nikon) microscope objective lens with the Lynx and the Tiger lasers, respectively.

 

Fig. 1. Schematic diagram of the polarized Raman microspectroscopic system illustrating laser excitation (blue solid line) and Raman signal detection (red dashed line) paths for acquiring tooth spectra in a backscattering sampling geometry. Abbreviations are provided in the inset for each optical element.

Download Full Size | PDF

Using the Lynx TEC-100 laser, good quality polarized Raman spectra could only be obtained after a total acquisition time of ~ 5 min per spectrum due to its low laser power. This long measurement time was undesirable because dehydration of the tooth could occur and lead to minor cracking of the sample. As a result, constant wetting of the tooth surface under measurement was therefore required. This problem could be solved by employing a higher power laser to reduce the total acquisition time. The higher power provided by the Tiger TEC-500 laser helped to reduce the total acquisition time to ~ 25 seconds per spectrum with similar spectral quality.

In order to measure the polarization dependence of Raman spectra of tooth enamel, a series of spectra were acquired at different angles (θ) of the polarization direction of the linearly polarized laser light with respect to that of the analyzer (PA). By mechanically rotating a half-wave plate placed in the laser path, laser polarization was varied in relation to that of the analyzer (PA). Figure 2 shows a schematic diagram on how the angle (θ) is determined. These measurements were carried out for sound enamel and carious lesions on the proximal surfaces of 4 unsectioned sample teeth collected from 3 patients. Subsequently, parallel-polarized and cross-polarized Raman spectra were obtained for each site from regions of sound enamel and identified carious lesions on the proximal surfaces of the 13 sample teeth. Parallel-polarized and cross-polarized spectra were obtained with the analyzer (PA) polarization oriented parallel (∥,θ=0°) and perpendicular (⊥,θ=90°) to that of the linearly polarized laser light, respectively. Spectra were acquired using 20 sec acquisition time and 15 accumulations for the Lynx laser or 5 sec acquisition time and 5 accumulations for the Tiger laser in order to obtain spectra with good signal-to-noise ratios. To examine how the laser probing direction relative to the tooth surface might affect the measurements, Raman spectra were collected with the tooth surface positioned approximately 30°, 45°, 60° or 90° relative to the laser propagation direction. Data were collected in triplicate from carious lesions and from sound enamel of 3 tooth samples at each probing angle.

 

Fig. 2. Definition of angle (θ): the angle between the laser polarization direction and the analyzer polarization direction (PA). HWP: half-wave plate

Download Full Size | PDF

For Raman imaging, spectral maps were acquired at 55 μm steps along both the x- and y-axes resulting in a 17 × 14 array map covering an area of 880 × 715 μm². A photomicrograph was taken of the area of the tooth surface that was mapped by Raman imaging using the same microscope objective. The tooth was side-illuminated using a fibre-optic white-light illuminator and carious lesions could be visualized as dark shadows. A ×5 microscope objective (Leica HC PL Fluotar) was used instead of the ×10 objective to provide larger field of view and enable a better visual comparison between adjacent carious lesions and surrounding enamel. Spectra used to produce the spectral images were acquired with both cross- and parallel-polarization states described earlier as ⊥ (θ=90°) and ∥ (θ=0°), respectively.

2.3 Data analysis

Spectral background subtraction and instrument response correction were performed according to the procedure outlined previously for both Raman point spectra and Raman spectral images [8]. Spectra were also corrected for the polarization dependence of the notch filter according to a procedure similar to that reported by Bremard et al. [18]

The depolarization ratio ρ₉₅₉ and polarization anisotropy A₉₅₉ were calculated according to conventional definitions:

where I₉₅₉(⊥) and I₉₅₉(∥) are the integrated peak intensities of the 959 cm^-1 peak detected with the analyzer oriented perpendicular to (⊥) and parallel to (∥) the polarization direction of the linearly polarized laser light, respectively. A depolarization ratio map was generated using MATLAB (The Mathworks, Inc., Natick, MA, USA) by dividing the cross-polarized (⊥) peak intensity map with the parallel-polarized (∥) peak intensity map. A corresponding anisotropy map was also generated using MATLAB for comparison. All spectral images were later smoothed using the bilinear interpolation function for 2-dimensional data in MATLAB.

For statistical analysis, ρ₉₅₉ values were calculated for 2 data groups, namely ρ_959(C) for carious lesions and ρ_959(S) for sound enamel. In order to determine if there is a significant difference between these two data groups, a t-test was carried out using the t-critical value at 99.9% confidence interval level with N₁+N₂-2 degrees of freedom; where N₁=32 and N₂=34 were the number of measurements for sound enamel and caries lesions, respectively, over 13 sample teeth. A similar statistical analysis was also performed for anisotropy, A_959(C) for carious lesions and A_959(S) for sound enamel. Spectroscopic data collected for the probing direction study were used to produce depolarization ratio and anisotropy values.

Sensitivity and specificity of the method were estimated using a Bayesian analysis model. Assuming a normal distribution, a threshold anisotropy value to distinguish sound enamel from carious enamel was estimated using normal distribution plots generated from experimentally determined mean and standard deviation values for carious and sound enamel. Each experimental anisotropy value was then compared with the threshold value to determine the true-positive (TP), false-positive (FP), false-negative (FN) and true-negative (TN) values. Sensitivity was calculated as TP/(TP+FN) and specificity was determined by TN/(TN+FP).

3. Results

Early carious lesions were not detected during the initial clinical assessment prior to extraction of the teeth and were not always visible on the dental radiographs. These lesions were detected after the teeth were extracted and the proximal surfaces were completely visible for clinical examination. The clinical assessments obtained ex vivo were used to guide the selection of regions for Raman spectroscopic measurements.

Representative polarized Raman spectra of sound enamel measured as a function of angle (θ) are shown in Fig. 3. Polarization dependence is evident with Raman peaks at 590 cm^-1, 608 cm^-1, 959 cm^-1, 1069 cm^-1 and 1104 cm^-1 as a function of angle (θ). These peak intensities decrease continuously as angle (θ) increases. Most noticeable is the intensity change of the 959 cm^-1 peak upon variation of laser polarization direction. This peak at 959 cm-1 represents the totally symmetric P-O stretching vibration of the phosphate ions (PO₄ ³) within the hydroxyapatite crystal of tooth enamel. All of the peaks experiencing polarization dependence have been previously assigned by both Tsuda et al [16] and LeRoy et al [17] to the A_g symmetry species of the phosphate ion (PO₄ ³) internal vibration modes of enamel crystallites. The exception is the 1104 cm^-1 peak, which arises from A-type carbonate ion (CO₃ ²) totally symmetric internal vibration. Similar polarization dependence is also present in spectra of carious lesions (Fig. 4), however, the amount of peak intensity change was not to the same extent. To describe such difference in a more quantitative measure, the integrated peak intensity at 959 cm^-1 obtained at each angle (θ) was first normalized to the peak intensity at θ=0° and was plotted as a function of angle (θ) for sound enamel and carious lesions —Fig. 5. Both traces show intensity minima at θ = 90° where the polarization direction of the linearly polarized laser light is oriented perpendicular (⊥) to that of the analyzer (PA). The 959 cm^-1 integrated peak intensity of carious lesions show lower sensitivity toward the polarization state of the incident laser.

 

Fig. 3. Representative polarized Raman spectra of sound enamel as a function of angle (θ). Inset (upper-left): enlargement of spectral region between 550 cm^-1 and 650 cm^-1. Inset (upper-right): enlargement of spectral region between 1000 cm^-1 and 1150 cm^-1. Angle (θ) was defined as the angle between the laser polarization direction and the analyzer polarization direction, as depicted in Fig. 2.

Download Full Size | PDF

 

Fig. 4. Representative polarized Raman spectra of carious lesion as a function of angle (θ). Inset (upper-left): enlargement of spectral region between 550 cm^-1 and 650 cm^-1. Inset (upper-right): enlargement of spectral region between 1000 cm^-1 and 1150 cm^-1. Angle (θ) was defined as the angle between the laser polarization direction and the analyzer polarization direction, as depicted in Fig. 2.

Download Full Size | PDF

In order to evaluate the reliability of using ρ₉₅₉ and A₉₅₉ for discriminating early carious lesion from sound enamel, a total of 66 ρ₉₅₉ values and 66 A₉₅₉ values consisting of 34 measurements from sound enamel and 32 measurements from carious lesions were examined over the 13 sample teeth collection. Figures 6(a) and 6(b) illustrate the distribution of ρ₉₅₉ and A₉₅₉ for both sound and carious enamel in bar-graphs. Statistical analyses resulted in a mean ρ_959(S) of 0.10±0.04 and a mean A_959(S) of 0.75±0.08 for sound enamel, a mean ρ_959(C) of 0.40±0.12 and a mean A_959(C) of 0.34±0.11 for carious lesions. A t-test performed at 99.9% confidence interval indicates that the differences in ρ₉₅₉ and A₉₅₉ between sound enamel and carious lesion are statistically significant with p < 0.001 for both ρ₉₅₉ and A₉₅₉. For estimating the sensitivity and specificity of the proposed methodology, experimental anisotropy data were compared with the threshold anisotropy value for distinguishing sound enamel from carious enamel. Of the measurements obtained from clinically identified carious lesions, only 1 data point was identified as sound enamel, i.e. 1 false-negative was identified. No false-positives were identified for the sound enamel tests. This Bayesian analysis results in a sensitivity of 97% and specificity of 100% for the proposed methodology based on Raman polarization anisotropy.

 

Fig. 5. Normalized integrated peak intensity at 959 cm^-1 measured as a function of angle (θ) for sound enamel (n=5) and carious lesion (n=5). Angle (θ) was defined as the angle between the laser polarization direction and the analyzer polarization direction, as depicted in Fig. 2.

Download Full Size | PDF

 

Fig. 6. Bar graphs of (A) depolarization ratio (ρ₅₉₅) and (B) polarization anisotropy (A₉₅₉) obtained from sound enamel versus carious lesion. Mean +/- standard deviation values are shown. n=34 and n=32 for sound enamel and carious lesion, respectively. For sound enamel, ρ_959(S)=0.10±0.04 and A_959(S)= 0.75±0.08 and for carious lesion, ρ_595(S)=0.40±0.12 and A_959(C)=0.34±0.11.

Download Full Size | PDF

Figure 7 provides an example of how the depolarization ratio or polarization anisotropy can be used to detect early carious lesions. Four Raman maps: a cross-polarized intensity map, a parallel-polarized intensity map, a depolarization ratio map and an anisotropy map are illustrated in Fig. 7 along with a photomicrograph of the tooth surface. The photomicrograph acquired under white light side-illumination shows two early carious lesions appearing as darker shadows when compared to neighbouring sound enamel. Dark shadow appearance of carious lesions under trans-illumination is well documented in the literature [19,20] and is believed to result from the poorer light transmission within the lesion. Although the sample tooth was side-illuminated instead of trans-illuminated (where the light source is placed at the opposite side of the tooth surface from which images are acquired), the same dark shadow appearance of carious lesion was still observed. A similar side-illumination method was recently demonstrated to be useful for a NIR imaging study on occlusal caries detection [11]. Among the Raman maps, both the depolarization ratio map and the anisotropy map (Fig. 7(c) and 7(d)) show high level of resemblance to the photomicrograph. Although Fig. 7(a), the cross-polarized intensity map, accurately detects the lesion locations and sizes, it however shows a region of “false-positive” at the upper-left corner of the mapped area. From Fig. 7(c) and 7(d), it is also clear that the degree of demineralization is greater at the centre than at the edge of the lesions. This same information is also present in the photomicrograph, however, the visual contrast in the Raman maps is clearer than that presented in the photomicrograph.

All the results presented so far are derived from data collected using a 90° laser probing angle with the laser beam perpendicular to the tooth surface. However, in reality, it would be difficult to ensure that the probing angle is at 90° for all the clinical caries assessments. Therefore, in order for this new method to be a clinically practical dental tool, it is essential that the depolarization ratio or polarization anisotropy results are reasonably independent of slight variations in the probing angle of the laser beam direction relative to the tooth surface. Polarized Raman spectra of sound enamel and carious lesions found on 3 sample teeth were acquired at various probing angles and values of ρ₉₅₉ and values of A₉₅₉ for sound enamel and carious lesions at each probing angle were subsequently determined. The mean±standard deviation (s.d.) values are presented in Fig. 8. Although a small dependence on probing angle was observed for both ρ₉₅₉ and A₉₅₉ values, the difference between sound enamel and carious lesions remained distinct at all angles. While data collected at 90° probing angle showed the highest degree of precision, no significant degradation in data precision was observed at all other angles. In order to measure the intra-sample reliability, another test measuring multiple locations on carious lesion and sound enamel on the same tooth was carried out. Despite some fluctuation in actual peak intensity, probing angle variation of 45° from the optimal 90° only slightly reduced the data contrast in both ρ₉₅₉ and A₉₅₉ values between sound enamel and carious lesions (data not shown for brevity).

 

Fig. 7. Raman spectral images of a tooth surface containing 2 carious lesions obtained using (a) cross-polarized (b) parallel-polarized peak intensities (c) depolarization ratio (d) anisotropy at 959 cm^-1. (e) photomicrograph of the same tooth surface under white-light side-illumination, showing 2 carious lesions as dark shadows. Area enclosed by solid black line shown in (e) indicates the region been mapped by Raman. The length of the scale bar shown at the corner of (e) is equivalent to 200 μm.

Download Full Size | PDF

4. Discussion

It is known that molecular alignment may cause polarization anisotropy of certain bands in the Raman spectrum therefore Raman spectra collected on well-aligned molecular structure at different polarization angles are expected to show substantial spectral variations. Since sound enamel is composed of carbonated hydroxyapatite crystals bundled in a highly ordered structure, its Raman spectral profile should vary as a function of polarization angle used to collect the Raman scattered light. As shown in Fig. 3 and Fig. 4, Raman peak intensities at 590 cm^-1, 608 cm^-1, 959 cm^-1, 1069 cm^-1 and 1104 cm^-1 all change as the initial laser polarization relative to analyzer polarization direction is varied. Partial loss of such Raman polarization anisotropy as observed in the spectra of carious lesions (Fig. 4) indicates possible alteration of an initially highly ordered molecular structure. This result supports the hypothesis reported in our previous study [8] that the non-polarized Raman spectroscopic changes detected on carious lesions were possibly due to alterations in the degree of enamel crystallite orientation during the demineralization process associated with caries formation.

 

Fig. 8. Depolarization ratio (ρ₅₉₅) and polarization anisotropy (A₅₉₅) obtained from sound enamel and carious lesions using 30°,45°,60°,90° laser beam probing angles relative to the tooth surface. Mean ± standard deviation (n=3) were shown for each angle based on data collected on sample teeth from different patients.

Download Full Size | PDF

A similar phenomenon was also observed with studies using polarization-sensitive optical coherence tomography (PS-OCT). For example, in studying the optical properties of carious lesions using PS-OCT, Baumgartner et al [21] observed a large amount of polarization phase change of the backscattered light from carious lesions. This change was not observed from sound enamel. As the tooth structure eventually depolarizes the incident polarized light through multiple scattering, the degree of phase change increases with the depth of light penetration. Therefore the backscattered light originating from the tooth surface should only show minimum polarization phase change. However, a larger amount of phase change was detected on a surface containing a carious lesion compared to sound enamel. Baumgartner et al have attributed this increased phase change to the alteration of birefringence due to increased scattering within carious lesions and/or alteration of enamel crystal orientation developed during the caries process. Combining the observation in this study, our previous study using non-polarized Raman spectroscopy [8] and the PS-OCT study [21], we believe that the cause for loss of Raman polarization anisotropy as observed in the spectra of carious lesions is a combination of increased scattering and structural alteration within the carious lesion. Based on information collected in the current study, we suggest that changes in Raman polarization anisotropy can be used as a marker for detecting the presence of early dental caries.

In order to utilize the changes in Raman polarization anisotropy for differentiating carious lesions from sound enamel, one easy way is to compare the cross-polarized spectrum with the parallel-polarized spectrum. Our result indicates that such information can be obtained by simply monitoring the intensity change of the 959 cm^-1 peak under cross- and parallel-polarized conditions. By using the depolarization ratio ρ₉₅₉ or degree of anisotropy A₉₅₉, change in the Raman polarization anisotropy can be readily determined. Comparing ρ₉₅₉ or A₉₅₉ instead of comparing actual peak intensities also helps to avoid potential problems caused by instrumental variations and sampling heterogeneity. It is known that different teeth or even different locations on one tooth can have different Raman scattering efficiency due to bio-heterogeneity, therefore the measured Raman intensity could vary from person to person or even from tooth to tooth. In contrast, ρ₉₅₉ and A₉₅₉, as forms of peak ratios, respond to the underlying structure but not to the types of experimental variations mentioned earlier. An example of how calculating peak ratios is useful for correcting spectral/sample artifacts is illustrated in Fig. 7. If the assessment is solely based on the cross-polarized 959 cm^-1 peak intensity (Fig. 7(a)), it can be mistakenly concluded that there are carious lesions at the upper-left corner of the tooth surface being imaged where, in fact no lesions are observed on the photomicrograph (Fig. 7(e)). The depolarization ratio map (Fig. 7(c)), on the other hand, successfully eliminates such artifacts and accurately displays lesion locations, sizes and possibly severity. Similarly, differences found in polarization anisotropy, A₉₅₉, can also be used to discriminate early carious lesions from sound enamel, as illustrated in Fig. 7(d).

The data presented demonstrates that both ρ₉₅₉ and A₉₅₉ can be used to differentiate carious lesions from sound enamel. However, these 2 values do differ in certain aspects such as their correlation to the sample’s structure and the accuracy of the data set. First, although using ρ₉₅₉ accurately detects carious lesions, a value of A₉₅₉ in fact more directly reflects the underlying structure of the sample. For example, sound enamel has a higher degree of molecular alignment. As a result it shows a higher degree of anisotropy (A₉₅₉). On the contrary, depolarization ratio is lower when higher degree of molecular alignment exists. Secondly, it is also observed that the anisotropy data points collected from sound enamel are less spread than those collected from carious lesions whereas the depolarization ratio data shows the opposite. Although the actual value of the standard deviation for the mean ρ_959(S) is smaller than that for the mean A_959(S), the relative standard deviation (the fraction of the actual value of standard deviation relative to the mean value) of the mean ρ_959(S) is much larger than that of the mean A_959(S). Statistically, the anisotropy data is believed to be a more rational parameter because the sound enamel sites should have similar structure and as a result, they should present with similar polarization anisotropy. In contrast, measured carious lesions are more likely to be at various stages of demineralization therefore it is reasonable to observe a wider range of polarization anisotropy values. Also, given the form of the denominator in the expression for anisotropy, it is a more robust measure compared to the simple depolarization ratio.

Earlier we mentioned that it is important to know whether variation of laser probing angle affects the test results. Both the intra- and inter-patient data presented shows a small variation in both ρ₉₅₉ and A₉₅₉ when the laser probing angle deviates from 90°, but the overall discriminating power of either ρ₉₅₉ and A₉₅₉ to detect early caries remains largely unaffected-see Fig. 8. This is especially true for probing angles between 45° and 90°.

In summary, to the best of our knowledge, this is the first time that polarized Raman spectroscopy is demonstrated as potential tool for characterizing early dental caries. Carious lesions can be detected based on a reduced Raman polarization anisotropy or a higher depolarization ratio of the 959 cm^-1 peak derived from polarized Raman spectra. The observed differences in polarization ratio or anisotropy is believed to be due to increased scattering and alteration in the degree of hydroxyapatite crystal orientation with demineralization. In addition, since ρ₉₅₉ or A₉₅₉ provides a single numerical value, this value can be used as a quantitative measure to assess the extent of demineralization. This parameter will not only be useful in aiding dentists in formulating dental treatment plans for their patients, but also provide a possible means to monitor a lesion for remineralization as a result of dental fluoride treatment. In order to better understand the potential and limitation of this proposed methodology, polarized Raman spectroscopic studies on artificial caries at various progressing stages and the effect of remineralization on enamel structure are currently under investigation.