2.1.

Specimen Preparation

A total of $155.0\text{-}\times 5.0\text{-}\times 2.0\text{-}\mathrm{mm}$ blocks were prepared from extracted bovine incisors. The initial enamel blocks were cut using a low-speed saw (ISOMET™ Low Speed Saw, Buehler, Ltd. Illinois). The dentin side was ground to establish a flat surface and the enamel side was ground and polished by RotoForce-4 and RotoPol-31 (Struers Inc., Rødovre, Denmark) until the total height was 2.0 mm. The specimens were mounted individually on 1-in. acrylic blocks using sticky wax, leaving only the polished enamel surface exposed.

2.2.

Baseline/Sound Surface Microhardness

The surface was divided into equal four areas, approximate size of $1\times 5{\mathrm{mm}}^2$. For each area, five baseline indentations were made using a Knoop diamond indenter (2100 HT; Wilson Instruments, Norwood, Massachusetts) with a 50-g load along a line parallel to the external surface of the specimen $\sim 100\mu \mathrm{m}$ apart from each other, with a dwelling time of 11 s. The Knoop hardness number for each specimen was derived by calculating the mean of the length of the long diagonal of the five indentations.

2.3.

Demineralization

The surface was checked under a stereomicroscope to make sure there was no debris on the enamel surface. Prior to the demineralization, the demineralizing solution²¹ with Carbopol 907™ (B.F. Goodrich Co. Chemical Group, Ohio) was warmed to 37°C. The solution contained $0.1\mathrm{mol}/\mathrm{L}$ lactic acid and 0.2% Carbopol 907™, was 50% saturated with hydroxyapatite and adjusted with a 0.25 N NaOH solution to pH 5.0. First, one area of enamel surface ($1\times 5{\mathrm{mm}}^2$) was covered with acid-resistant clear nail varnish as baseline prior to demineralization. Specimens were exposed to 12.4-mL demineralization solution for 8 h at 37°C. After 8-h demineralization (8-h demin), one area ($1\times 5{\mathrm{mm}}^2$) next to baseline was covered with acid-resistant clear nail varnish and the specimens were exposed to demineralization solution for additional 8 h at 37°C. After additional 8-h demineralization (total 16-h demineralization: 16-h demin), one area ($1\times 5{\mathrm{mm}}^2$) next to 8 h demineralized area (16-h demin area) was covered with acid-resistant clear nail varnish and the specimens were exposed to demineralization solution for another additional 8 h at 37°C. At the end, there were four areas on each surface: 0-, 8-, 16-, and 24-h demineralization (Fig. 1). The clear nail varnish was then removed with acetone.

Fig. 1

This diagram shows physical dimension of specimen and areas for demineralization. Black squares indicate the locations of SRS image acquisition and black circles indicate the locations of SRS measurements. The left side of image presents sound (0-h demineralization) on the surface. Darker color indicates longer demineralization time.

2.4.

Demineralized Enamel Surface Microhardness (${\mathrm{SMH}}_{\mathrm{D}}$)

After demineralization, a second set of five indentations was made using the Knoop diamond indenter as described in Sec. 2.2, but to the left of and parallel to the sound enamel indentations, $\sim 100\mu \mathrm{m}$ apart from each other and $\sim 200\mu \mathrm{m}$ from the sound enamel indentations. The Knoop hardness number for each specimen was derived by utilizing the mean of the length of the long diagonal of the five indentations. Surface microhardness-change (SMH-change) [$({\mathrm{SMH}}_{\mathrm{DN}}-{\mathrm{SMH}}_{\mathrm{S}})/{\mathrm{SMH}}_{\mathrm{S}}$] was determined. $N$ indicates demineralization time.

2.5.

Hyperspectral SRS Microscopy

Our hyperspectral SRS imaging system (Fig. 2) was built based on a previously reported spectral focusing scheme,²² where the Raman shift is controlled by the temporal delay of two chirped femtosecond pulses. An ultrafast laser system with dual output operating at 80 MHz (InSight DeepSee, Spectra-Physics) provided the excitation sources. The tunable output with a pulse duration of 120 fs was tuned to 942 nm, serving as the pump beam. The other output centered at 1040 nm with a pulse duration of 220 fs was used as the Stokes beam, modulated by an acousto-optic modulator (AOM, 1205-C, Isomet) at 2.3 MHz. An optical delay line was installed in the pump light path and a motorized translation stage (T-LS28E, Zaber) was employed to scan the delay between the two beams. The spectral coverage from $\sim 850$ to $1150{\mathrm{cm}}^{-1}$ was tested. After combination by a dichroic mirror, both beams were chirped by two 12.7-cm-long SF57 glass rods. To match the linear chirp parameter of the two beams, the built-in prism compressor inside the laser system was used to adjust the dispersion of the pump beam. After glass rods, pulse durations of the pump and Stokes beams were stretched to $\sim 0.8$ and $\sim 1.3\mathrm{ps}$, respectively, measured with an autocorrelator (PulseScope, APE). The pump and Stokes beams were sent into a home-built laser-scanning microscope. A $20\times$ water immersion objective lens ($\mathrm{NA}=1.0$, XLUMPLFLN-W, Olympus) was utilized to focus the light onto the sample. The back-scattered signal was collected by the same objective, reflected by a polarizing beam splitter to a lens, and then focused on a Si-photodiode (S3994-01, Hamamatsu). The SRS signal was extracted with a digital lock-in amplifier (HF2LI, Zurich Instrument).

Fig. 2

Epidetected hyperspectral SRS microscope. AOM, acousto-optic modulator; DM, dichroic mirror; F, filter; HWP, half waveplate; L, lens; OBJ, objective; PBS, polarizing beam splitter; and PD, photodiode.

2.6.

SRS Data Analyses

2.7.

Transverse Microradiography (TMR)

A thin section, $\sim 100\mu \mathrm{m}$ in thickness, was cut from the center area of each specimen using a Silverstone–Taylor Hard Tissue Microtome (Scientific Fabrications Laboratories). Any section thicker than $120\mu \mathrm{m}$ was hand-polished using 2400-grit silicon carbide paper to the required thickness. The sections were mounted with an aluminum step wedge on high-resolution glass plates type I A (Microchrome Technology Inc., San Jose, California). Sections were placed in the TMR-D system and $x$-rayed at 45 kV and 45 mA at a fixed distance for 12 s. The digital images were analyzed using the TMR software v.3.0.0.18 (Inspektor Research Systems BV, Amsterdam, the Netherlands). A window $\sim 400\times 400\mu \mathrm{m}$ representing the entire lesion and not containing any cracks, debris, or other alterations was selected for analysis. The following variables were recorded for each specimen: lesion depth ($L$) (87% mineral; i.e., 95% of the mineral content of sound enamel), and integrated mineral loss ($\mathrm{\Delta }Z$), which is calculated by subtracting the mineral profile in the lesion area from the sound profile.

2.8.

Statistical Analyses

Comparisons among demineralization times for differences in SMH-change, $L$, $\mathrm{\Delta }Z$, SRS P/C-ratio-normalized, SRS P/C depth, and SRS water content were made using repeated measures of analysis of variance (ANOVA). The ANOVA included a fixed effect for demineralization time, with demineralization time repeated with specimen, allowing each demineralization time to have a different variance and different covariances were allowed between times. A 5% significance level was used for all tests. Pearson correlation coefficients were calculated to evaluate the linear associations between measurements.

3.1.

SRS Images at Raman Bands of Phosphate, Carbonate, and Water

Figure 3 shows representative SRS images of phosphate, carbonate, and water Raman bands. An 8-h demin showed microscopic demineralized structures at phosphate and carbonate Raman bands [Figs. 3(a) and 3(b)] and higher water penetration than sound area [Fig. 3(c)]. Following by the same trend, 24-h demin showed lower SRS signal at phosphate and carbonate Raman bands and more demineralized indications than 16-h demin [Figs. 3(d) and 3(e)]. The water penetration in the 24-h demineralized region was higher than 16-h demin [Fig. 3(f)].

Fig. 3

SRS images of specimen at Raman bands of phosphate, carbonate, and water bands. These images were taken $30\mu \mathrm{m}$ below surface. (a–b) About 8-h demin showed demineralized structures than sound area and (c) higher water penetration. (d–e) About 24-h demin showed more demineralized structures than 16-h demin and (f) higher water penetration. The SRS spectra of (a–f) are shown in (g). Scale bar for all panels: $100\mu \mathrm{m}$.

3.2.

Observation of Demineralized Enamel by SMH, SRS, and TMR

Table 1 shows means and standard deviations of SMH-change, SRS P/C depth, $L$, and $\mathrm{\Delta }Z$. SMH-change increased as demineralization time increased; all four demineralization times were significantly different from each other ($p<0.001$). For SRS P/C depth, depth increased with demineralization time (sound versus 8-h demin $p=0.0086$; sound versus 16-h demin $p=0.0134$; sound, 8-h demin, and 16-h demin versus 24-h demin $p<0.0001$, 8-h demin versus 16-h demin $p=0.20$). For depth of lesion ($L$), depth increased significantly with demineralization time (8-h demin versus 16-h demin, $p=0.0112$; 8-h demin versus 24-h demin, $p=0.0001$; 16-h demin versus 24-h demin, $p=0.0004$). For integrated mineral loss ($\mathrm{\Delta }Z$), 24-h demin was significantly larger than 8-h demin ($p=0.0005$) and 16-h demin ($p<0.0001$), but 8-h demin and 16-h demin were not significantly different ($p=0.15$). There were moderate correlations between $L$ (lesion depth) and SRS P/C depth ($r=0.55$, $p<0.001$) and between $\mathrm{\Delta }Z$ (integrated mineral loss) and SRS P/C depth ($r=0.49$, $p=0.001$).

Table 1

Means±standard deviations of surface microhardness-change (SMH-change), SRS P/C ratio norm depth (SRS P/C depth), lesion depth (L), and integrated mineral loss (ΔZ) for each demineralization time.

3.3.

Evaluation of Demineralized Enamel by SRS P/C-Ratio-Normalized

3.3.1.

Mean and standard deviation

Table 2 shows means and standard deviations of SRS P/C Ratio relative to $90\text{-}\mu \mathrm{m}$ depth (SRS P/C-ratio-normalized) for each demineralization time at each depth; the means are shown in Fig. 4. Sound (0-h demineralization: 0-h demin) was significantly higher than 8-h demin at depths 0 to $10\mu \mathrm{m}$; but lower than 8-h demin at depths 40 to $80\mu \mathrm{m}$. 0-h demin was significantly higher than 16-h demin at depths 10 to $30\mu \mathrm{m}$; but lower than 16-h demin at depths 50 to $70\mu \mathrm{m}$. 0-h demin was significantly higher than 24-h demin at depths 0 to $40\mu \mathrm{m}$; but lower than 24-h demin at depths 60 to $70\mu \mathrm{m}$. 8-h demin was significantly higher than 16-h demin at depths 20 to $40\mu \mathrm{m}$; but significantly lower than 16-h demin at a depth of $70\mu \mathrm{m}$. An 8-h demin was significantly higher than 24-h demin at depths 10 to $40\mu \mathrm{m}$; but significantly lower than 24-h demin at the depth of $70\mu \mathrm{m}$. 16-h demin was significantly higher than 24-h demin at depths 0 to $40\mu \mathrm{m}$.

Table 2

Means±standard deviations of SRS P/C ratio relative to 90-μm depth (SRS P/C-ratio-normalized) for each demineralization time at each depth.

Depth (μm)	Demineralization time (h)
	Sound (0)	8	16	24
0 (Surface)	$0.66\pm 0.16$a1	$0.58\pm 0.17$b,c1	$0.65\pm 0.16$a,b1	$0.51\pm 0.16$c1
10	$0.99\pm 0.07$a2	$0.92\pm 0.12$b2	$0.90\pm 0.10$b2	$0.77\pm 0.11$c2
20	$1.05\pm 0.04$a3	$1.05\pm 0.08$a3–5	$0.98\pm 0.06$b3	$0.88\pm 0.08$c3
30	$1.05\pm 0.03$a4	$1.08\pm 0.06$a6	$1.02\pm 0.06$b4	$0.95\pm 0.08$c4
40	$1.04\pm 0.03$b5	$1.07\pm 0.04$a5,6	$1.04\pm 0.04$b5	$1.01\pm 0.06$c5
50	$1.03\pm 0.02$b6	$1.05\pm 0.03$a4	$1.04\pm 0.04$a5	$1.04\pm 0.06$ab6
60	$1.01\pm 0.03$b2	$1.03\pm 0.03$a3	$1.04\pm 0.03$a4,5	$1.05\pm 0.06$a6
70	$1.01\pm 0.02$c2	$1.02\pm 0.02$b3	$1.04\pm 0.03$a4,5	$1.05\pm 0.05$a6
80	$1.01\pm 0.02$a2	$1.01\pm 0.02$b3	$1.03\pm 0.04$a,b4,5	$1.04\pm 0.05$ab5,6

Same letters indicate no significant differences among demineralization times at each depth (horizontal comparison) and same numbers indicate no significant differences among depths within each demineralization time (vertical comparison).

The superscript numbers indicate no significant differences among the depths within each demineralization time (vertical comparison).

Fig. 4

Means of SRS P/C ratio relative to $90\text{-}\mu \mathrm{m}$ depth (SRS P/C-ratio-normalized) for each demineralization time at each depth.

3.4.

Quantification of Demineralized Enamel by Water Content Measured by SRS

3.4.1.

Mean and standard deviation

Figure 5 shows water Content measured by SRS. About 24-h demin had significantly higher water content than all other groups ($p<0.001$ versus 0 h, $p=0.002$ versus 8 h, $p=0.001$ versus 16 h), and 8-h demin and 16-h demin had significantly higher water content than 0-h demin ($p<0.001$), but 8-h demin and 16-h demin were not significantly different from each other ($p=0.84$).

Fig. 5

$\text{Means}\pm \text{standard deviations}$ of SRS water content for each demineralization time. Same letters indicate there is no significant differences.

3.4.3.

3-D Image of water content by SRS

Figure 6 shows representative three-dimensional (3-D) images for SRS Water Content. “a” shows area between sound (0-h demineralization) and 8-h demineralization. “b” shows area between 16-h demineralization and 24-h demineralization. The left side of each image is the shorter demineralization time. Darker color (blue) indicates more water in the enamel. As demineralization time increased, the area became darker.

Fig. 6

Example images for SRS water content. (a) Area between sound (0-h demineralization) and 8-h demineralization. (b) Area between 16-h demineralization and 24-h demineralization. The left side of each image presents the shorter demineralization time. Darker color (blue) indicates more water in the enamel.

4.1.

SRS Potentially Can Quantify Demineralized Enamel

Human enamel has attained 86% mineral by volume.¹⁵ The mineral component of sound enamel consists of carbonated calcium hydroxyapatite crystals [${\mathrm{Ca}}_{10}({\mathrm{PO}}_4{)}_6{\mathrm{OH}}_2$].^16,17 Extraneous ions such as carbonate, fluoride, sodium, and magnesium are also frequently found within the crystal structure. Sound enamel contains 2% to 4% carbonate (by weight) and 17.5% to 18.5% phosphate.^17,20,24,25 Dental caries is demineralization of dental hard tissue, which is the loss of calcified material, such as calcium and phosphate, from the tooth structure. This also appears to be complemented by preferential decrease in carbonate.^19,26 Lesion depth and mineral loss were increased linearly against demineralization time.²⁷ The rate of demineralization calculated from the titration data showed also nearly linear increase with time.²⁸ Relative loss of carbonate ions occurs in the first minutes of the reaction at pH 5.6.²⁹ Mineral removal/loss of the caries lesion compared with sound enamel was calculated.²⁰ From the lesion body, phosphate removed/lost was 18.5% and carbonate was 1.0%. Based on these data, although carbonate decreased rapidly, nearly 99% of carbonate might remain in the lesion body. Thus, when the amount/intensity of phosphate is divided by that of carbonate (P/C ratio), the P/C ratio should decrease as the severity of the lesion/demineralization increases. For our current study, to quantify and compare severity of demineralization among different teeth/specimens, peak ratio was used and defined as the phosphate peak height divided by the carbonate peak height (SRS P/C ratio). SRS P/C ratio was further normalized relative to $90\text{-}\mu \mathrm{m}$ depth (SRS P/C-ratio-normalized). The results showed at the surface level ($0\text{-}\mu \mathrm{m}$ depth in Table 2) that SRS P/C-ratio-normalized decreased as demineralization time increased. There was a significant difference between sound and 24-h demineralized enamel, which had an average depth of $50.7\mu \mathrm{m}$. Previous studies demonstrated that Raman spectroscopy could detect dental caries.^9–11 The principle of these studies is that the Raman peak at $959{\mathrm{cm}}^{-1}$ (phosphate) for caries is weaker than that for sound enamel. Another study observed a cross-sectioned surface of the tooth.¹⁴ This study indicated that SRS could quantify caries lesions using the Raman peak at $959{\mathrm{cm}}^{-1}$. Detection of caries lesions is the first step and an important process for caries management/treatment. To determine an appropriate treatment/management modality, quantification of the severity of caries lesion/demineralization is also important because severe lesions may require more intense treatment/management. Our study successfully showed that SRS has the potential to quantify and visualize caries lesion/demineralization based on the intrinsic phosphate and carbonate contents. Similarly, the study of SRS polarization/depolarization has been reported to identify dental lesions.¹³ In the future, the combination of SRS polarization/depolarization and hyperspectral measurements could potentially provide more information of demineralization on the molecular level. It is well known that although sound enamel is highly mineralized tissue, the chemical composition is not uniform from the surface to inner structure.³⁰ From the surface to the internal of the enamel, calcium, phosphate, and fluoride concentrations seem to decrease. On the other hand, magnesium, carbonate, and chloride concentrations seem to increase. However, concentration differences varied from place to place in the tooth. In addition, this variation may depend on the distance from the surface.³¹ These are the potential reasons for SRS P/C ratio changes from the surface to inner structure for the sound enamel.

4.2.

SRS Can Potentially Estimate the Depth of Demineralized Enamel

Another unique aspect of our current study is that SRS P/C-ratio-normalized data can potentially estimate the depth of the lesion/extent of demineralization (SRS P/C depth). Although the correlation was moderate, this indicates that SRS should be able to estimate depth of lesions from the anatomical surface without sectioning the teeth/specimens, which would not be possible under clinical conditions. The use of $10\text{-}\mu \mathrm{m}$ increments at which the SRS measurements were performed would decrease the precision of the SRS P/C depth assessments. This may have lead to a reduction in the correlations with TMR lesion depth and $\mathrm{\Delta }Z$. Further studies will need to confirm the present findings and improve the methodology.

4.3.

SRS-Based Water Content Can be Used to Determine Degree of Demineralization

Our current study is the first report investigating the relationship between the amount of water and the severity of caries lesion/demineralization. At eruption, human enamel has attained 86% mineral, 2% organic material, and 12% water by volume.¹⁷ The mineral component of sound enamel consists of tightly packed hydroxyapatite crystals. These tightly packed hydroxyapatite crystals are separated by tiny intercrystalline spaces. Based on chemical analysis and histopathological observations, the initial stage of caries development is characterized by the opening of the intercrystalline spaces without the destruction of the surface and subsequent creation of microchannels.^32–34 This initial stage of caries lesion, that macroscopically maintains intact/sound surface, is called a noncavitated caries lesion. As demineralization continues, while maintaining intact surface, microchannels get wider and longer³⁵ and the lesion body becomes more porous.³⁶ When noncavitated caries lesions are hydrated/wet, microchannels and the lesion body are filled with saliva/fluid. It is reasonable to assume that deeper noncavitated lesions contain more water, as demineralization continues. The results of our current study showed that as demineralization time increased, the SRS signal at water band increased. There were good correlations between the SRS signal at water band and the hardness and microradiography variables. We also successfully presented 3-D imaging of water from the top surface without sample preparation. As demineralization time increased, the signal from water got stronger, indicating higher penetration of water into demineralized enamel. Our observation is also consistent with previous work on measurement of refractive indices of sound and demineralized enamel. According to Refs. 3738.–39, the refractive index of sound enamel is 1.62, and interestingly the refractive index demineralized enamel is reported as 1.35,³⁹ which is close to water refractive index (1.33). The lower refractive index of demineralized enamel is possibly due to the water penetration.

We note that in demineralized enamel, the porous structure filled with either air or water exhibits lower refractive index compared with sound enamel. This could lead to wavefront distortion and multiple optical scattering, which can degrade SRS signal at water band. As a result, water contained in demineralized enamel might have been under-estimated. Future study will be required to quantified the water content in demineralized enamel. Despite this concern, enamel regions with longer demineralized time generally exhibit stronger SRS water signal [Figs. 3(c) and 3(f)], which supports our hypothesis. Light scattering can also degrade the Raman signals of phosphate and carbonate groups. In this study, we used the peak ratio between phosphate and carbonate groups to quantify enamel demineralization. Although light scattering can reduce the light intensity delivered into lesions and also degrade the collection of Raman signals, the ratio between two Raman peaks remains a quantifiable measurement. Moreover, the advantage of SRS measurement includes the use of long wavelength in the near infrared region, which enables us to penetrate hundreds of micrometers. Also as one of the nonlinear optical techniques, SRS is advantageous in blocking out-of-focus light compared with other linear techniques.

4.4.

Comparing with other Technology-Based Methods

There are other optical caries detection methods available, such as reflectance-based method,⁴⁰ optical coherence tomography (OCT),⁴¹ pulsed heating thermography (PHT),⁴² and light-induced fluorescence (LF).⁶ OCT measures the changes in reflectivity or light scattering with millimeter penetration depth. PHT provides heat maps that could be used to localize infection or inflammation. LF uses short wavelength, usually blue light, to induce autofluorescence of tooth structure, especially of dentin. These optical modalities show advantages of nondestructive and quantitative assessment and eliminating the need of ionizing radiation. OCT, PHT, and LF are quantitative methods that can determine the severity of caries lesions, such as depth or equivalent of depth of the lesions. One of the limitations for OCT can be dental structures’ images, which can be acquired up to 2.5 mm in depth.⁴³ Near-infrared light transillumination (NILT)^41,44 has also been reported with the advantage of being able to be used on the approximal surface, which remains a difficult measurement for other modalities. One of the disadvantages of NILT can be currently there is no quantification function to determine severity of caries lesions. Among these available optical methods, the fundamental principle is to detect/determine physical damage/destruction of tissue structure. On the other hand, Raman spectroscopy is not only a nondestructive method, but also a chemical analysis method that provides detailed information about the chemical structure, phase and polymorphy, crystallinity and molecular interactions, which cannot be accessed by other modalities. SRS provides a unique approach to visualize the chemical composition based on their molecular vibrations. In this study, we reported the direct measurement of increased water content in demineralized enamel with submicron spatial resolution, which provides a potential marker to quantify demineralization of enamel. We also reported the use of two inherent Raman peaks from enamel to quantify the demineralization. The ratio of phosphate and carbonate groups enables quantification of demineralization on the molecular level. Many optical methods, such as OCT, determine the optical changes to determine severity of caries lesions, such as depth of lesions. SRS, on the other hand, obtains chemical composition and is able to provide analysis of demineralization on the molecular level. However, SRS is mostly used in a benchtop-base setting and primarily for the laboratory use. In the future, we expect that more developments in handheld SRS microscope or endoscope will be reported.⁴⁵