1.

Introduction

New, more sophisticated diagnostic tools are needed for the detection and characterization of caries lesions in the early stages of development. If carious lesions are detected early enough, they can be arrested/reversed by nonsurgical means through fluoride therapy, antibacterial therapy, dietary changes, or by low intensity laser irradiation.

Dentists also need an imaging technology that can reliably track the course of caries lesions over an extended time period to determine whether the lesion is active and expanding or whether the lesion has been arrested, to determine if intervention is needed. In our laboratory, we are developing near-IR imaging methods for caries detection that exploit changes in the light scattering properties of dental enamel that occur on demineralization.^{1, 2}

Dental enamel manifests high transparency in the near-IR, therefore the near-IR is ideally suited for: transillumination of teeth for the detection of interproximal caries,² near-IR imaging of occlusal caries,³ and for imaging with optical tomographic methods such as optical coherence tomography.⁴ To effectively develop such technologies and optimize their effectiveness, it is necessary to understand how near-IR light propagates through the tooth and how the optical properties of the tooth change as a result of natural demineralization or tooth decay. The aim of this work was to determine how the scattering properties of dental enamel change with natural and artificial demineralization.

The principal factor limiting caries detection via optical imaging methods in the visible range of 400 to $700\mathrm{nm}$ is light scattering. Light scattering in sound enamel and dentin is sufficiently strong in the visible range to obscure light transmission through the tooth. The magnitude of light scattering in dental enamel is expected to decrease $(1∕{\lambda }^3)$ due to the size of the principal scatterers in enamel. Therefore, the near-IR region from 780 to $1550\mathrm{nm}$ offers the greatest potential for new optical imaging modalities due to the weak scattering and absorption in dental hard tissue.⁵ Ballistic transmission measurements have demonstrated that enamel has the highest transparency in the near-IR region near $1310\mathrm{nm}$ .⁶ At this wavelength, the attenuation coefficient is only 2 to $3{\mathrm{cm}}^{-1}$ , a factor of 20 to 30 times lower than in the visible region. At longer wavelengths, water absorption increases markedly, reducing the penetration of IR light (see Fig. 1 ).

Fig. 1

Plot of the extinction (attenuation) coefficient of dental enamel (filled circles) and the absorption coefficient of water (open circles) versus wavelength.^{5, 6, 30}

Dental hard tissue optics is inherently complex due to the nonhomogeneous and anisotropic nature of these biological materials. Dental enamel is an ordered array of inorganic apatite-like crystals surrounded by a protein/lipid/water matrix. The crystals are approximately 30 to $40\mathrm{nm}$ in diameter and can be as long as $10\mathrm{\mu }\mathrm{m}$ . The crystals are clustered together in $4\text{-}\mathrm{\mu }\mathrm{m}$ -diameter rods (or prisms), which are roughly perpendicular to the tooth surface. Because of the complex nature of these materials, the scattering distributions are generally anisotropic and depend on tissue orientation relative to the irradiating light source, ^{5, 7, 8, 9} in addition to the polarization of the incident light.

The optical parameters for normal enamel and dentin have been reported between the wavelength range of 200 to $700\mathrm{nm}$ . For enamel, absorption is very weak in the visible range ( ${\mu }_a<1{\mathrm{cm}}^{-l}$ , $\lambda =400$ to $700\mathrm{nm}$ ) and increases in the ultraviolet $({\mu }_a>10{\mathrm{cm}}^{-1},\lambda <240\mathrm{nm})$ .¹⁰ For dentin in the 400 to $700\mathrm{nm}$ wavelength range, the absorption coefficient is essentially wavelength independent with a value of ${\mu }_a\sim 4{\mathrm{cm}}^{-1}$ .¹¹ The optical properties of normal enamel and dentin, including the scattering phase function and anisotropy, were determined at 1053, 632, and $543\mathrm{nm}$ using angular resolved light scattering measurements and Monte Carlo (MC) simulations of light propagation.⁵

The optical properties of biological tissue can be quantitatively described by defining the optical constants, the absorption $\left({\mu }_a\right)$ and scattering coefficients $\left({\mu }_s\right)$ , which represent the probability of the incident photons being absorbed or scattered, and the scattering phase function $\Phi \left[\cos \right(\theta \left)\right]$ , which is a mathematical function that describes the directional nature of scattering.^{12, 13, 14} With knowledge of these parameters, light transport in dental hard tissue can be completely characterized and modeled. Accurate description of light transport in dental hard tissue relies on knowledge of the exact form of the phase function $\Phi \left[\cos \right(\theta \left)\right]$ for each tissue scatterer at each wavelength.¹⁴ Therefore, direct measurement of the phase function is necessary. It is important to note that the empirically derived Kubelka-Munk (KM) coefficients that are commonly used in dentistry are not normal fundamental optical constants and are not appropriate for describing light transport in tissues with highly forward directed scattering such as enamel and dentin, particularly in the near-IR. ^{12, 13, 14, 15} Van der Zee¹⁵ compared KM methods with Monte Carlo methods and found that the KM method led to considerable errors and that it was necessary to use the more rigorous MC method for accurate results.

We demonstrated previously that angular resolved scattering distributions of enamel and dentin representing single scattering could be measured. This was experimentally achieved by using thin sections with a thickness much less than the photon mean free path (MFP), to ensure single scattering, and an index-matching bath to remove the influence of surface scattering and refractive effects. The optical properties were subsequently determined using Monte Carlo simulations of photon transport. Although this approach is laborious and computationally intensive, it has been shown consistently to yield the most reliable values.^{12, 14} The accuracy of the values computed was confirmed by successfully modeling the angular resolved scattering distributions measured for sections of various thickness from $100\mathrm{\mu }\mathrm{m}$ to $2\mathrm{mm}$ in the multiple scattering regime. Applying this method, we were able to accurately determine the optical constants and the scattering phase function for sound enamel and dentin at 543, 632, and $1053\mathrm{nm}$ .^{5, 16} More recently, we measured the attenuation coefficient ${\mu }_t$ , which is the sum of the scattering and absorption coefficients at 1310 and $1550\mathrm{nm}$ .⁶ Measurement of the enamel attenuation coefficient at 1310 and $1550\mathrm{nm}$ posed a particular challenge, since the scattering coefficient is almost two orders of magnitude lower than in the visible range, $\sim 3{\mathrm{cm}}^{-1}$ at $1310\mathrm{nm}$ versus $105{\mathrm{cm}}^{-1}$ at $543\mathrm{nm}$ . Moreover, surface scattering can completely mask the measurement of the bulk scattering properties.⁵

The measured angular resolved scattering distributions could not be represented by a single scattering phase function $\Phi \left(\cos \theta \right)$ , and required a linear combination of a highly forward peaked phase function, a Henyey-Greenstein (HG) function, and an isotropic phase function represented by the following equation:⁵

The fraction of isotropic scatter $\left(f_d\right)$ was measured to be 36% for enamel and less than 2% for dentin at $1053\mathrm{nm}$ . Due to the heterogeneous structure of dental hard tissues, the phase function changes with wavelength; therefore, the phase function has to be independently determined at each wavelength. Angular resolved scattering measurements on a 2-mm-thick section of sound enamel at $1310\mathrm{nm}$ previously coupled with Monte Carlo simulations yielded a HG function with $g=0.96$ and $f=55\%$ isotropic scattering.¹⁷ The scattering coefficient used in those simulations was $3{\mathrm{cm}}^{-1}$ . It is interesting to note that the shape of the scattering distributions measured at $1310\mathrm{nm}$ is similar to those measured at $1053\mathrm{nm}$ . with the exception of the markedly higher dynamic range $\sim 100\mathrm{db}$ , and minimal reduction in the ballistic light transmitted at $0\mathrm{deg}$ . For comparison, at $543\mathrm{nm}$ the ballistic component at $0\mathrm{deg}$ was no longer visible in sections greater than $1\mathrm{mm}$ thickness, and the diffuse (multiply scattered component) at angles greater than 5 to $10\mathrm{deg}$ was of similar magnitude. At $1310\mathrm{nm}$ , the ratio of intensity of the unscattered component is still $50\mathrm{dB}$ or five orders of magnitude higher than the diffuse component for the sample of $2\mathrm{mm}$ thickness.

The nature of the scattering phase function suggests that there are two types of scatterers in enamel and dentin. We hypothesize that the small enamel crystallites of less than $50\mathrm{nm}$ in diameter produce Rayleigh-like isotropic distributions and are responsible for the large fraction of isotropic scattering from the tissue. The large enamel prisms and dentinal tubules that are a few microns across are Mie-like scatterers, producing highly forward directed scattering distributions that can be described by a HG function, the second term in Eq. 1. We also postulate that demineralization will substantially increase the fraction of isotropic scatterers $\left(f_d\right)$ and increase the amount of large angle scattering from the carious lesion. To test this hypothesis, the angular distribution of scattered light was measured in thin sections of enamel exposed to a pH cycling regimen to produce artificial or simulated caries lesions.

The first step to determining the optical properties of carious enamel is to establish a method of characterizing the severity of the carious lesion. The gold standard in cariology research is mineral loss, which defines the degree of demineralization in the lesion. The most severe lesions have lost a higher volume fraction or volume percent of their original mineral content than less severe lesions. We have assembled, tested, and calibrated a high-resolution digital microradiography (DM) system—spatial resolution $2.1\mathrm{\mu }\mathrm{m}$ —for quantification of lesion severity (Fig. 2 ). The DM system allows real-time acquisition of digital x-rays and works well for thicker sections, greater than the $80\text{-}\mathrm{\mu }\mathrm{m}$ thin sections typically required for conventional film-based microradiography. This is a major advantage, since it is difficult to cut $80\text{-}\mathrm{\mu }\mathrm{m}$ thin sections without destroying natural lesions, and grinding and polishing modifies the soft lesion.

Fig. 2

High-resolution digital x-ray microradiography system (DM), (a) FDI x-ray imager with motion control system, (b) x-ray diffractometer, and (c) sample mount.

Increased backscattering from the demineralized region of early caries lesions is the basis for the visual appearance of white spot lesions.^{18, 19} An increase in porosity of the lesion leads to increased scattering at the lesion surface and higher scattering in the body of the lesion, producing an increase in the magnitude of the diffuse reflectance.²⁰ Attempts at measuring the optical properties of dental caries have been limited to measurements of backscattered light from optically thick, multilayered sections of simulated caries lesions.^{20, 21, 22} The microstructure of enamel and dentin caries lesions is complex, consisting of various turbid and transparent zones.^{23, 24} Demineralized areas of coronal caries appear whiter and are more opaque. During remineralization, pores and tubules are filled with minerals, and those areas are typically more transparent. There have been some attempts to measure and simulate light scattering in artificial and natural lesions.^{18, 22, 25} However, to our knowledge there has not been a major effort to measure and quantify the optical properties of carious lesions in terms of changes in the fundamental optical constants ${\mu }_s$ and ${\mu }_a$ and the scattering phase function. In this study, we have used lasers, precision positioning stages, and a zoom video microscope system to accurately position the laser beam with a spot size of 50 to $100\mathrm{\mu }\mathrm{m}$ on various areas of carious lesions for direct comparison with matching acquired DM images. One procedural difference between measuring carious lesion areas versus sound tissue areas is the method of index matching. Lesion areas are highly porous, allowing index matching fluids to imbibe into the lesion, effectively eliminating the lesion from an optical standpoint, i.e., the lesion becomes more transparent than the sound tissue. Our method is predicated on the principle that carious lesions will be filled with fluid in their natural state, and that measurement in water replicates the conditions that will be encountered clinically.

2.

Materials and Methods

2.2.

Simulated-Caries Lesions

Simulated caries lesions were produced in $200\text{-}\mathrm{\mu }\mathrm{m}$ -thick human tooth sections using well-characterized demineralizing and remineralizing solutions for varying time periods to produce simulated caries lesions of varying severity. The pH cycling model that replicates the cycle of demineralization and remineralization that takes place naturally in the mouth as the pH fluctuates before and after meals^{26, 27} was used. Both sides of human tooth sections were exposed for 1, 2, 3, 4, and 5 days to a daily regimen of 6-h demineralization by 17-h remineralization (see Fig. 3 ). A total of five sections were used and the scattering distributions were measured after each full daily period of pH cycling. In the pH cycling model, a surface zone is created due to the protection of the surface layer, and erosion does not occur as the lesion progresses in depth. Each section was immersed in 40-mL aliquots of a buffer solution containing 2.0-mmol/L calcium, 2.0-mmol/L phosphate, and 0.075-mol/L acetate maintained at pH 4.3 and a temperature of $37°\mathrm{C}$ for a period of $6\mathrm{h}$ each day, followed by immersion in a remineralization solution overnight for $17\mathrm{h}$ in a solution of 1.5-mmol/L calcium, 0.9-mmol/L phosphate, 150-mmol/L $\mathrm{KCl}$ , and 20-mmol/L cacodylate buffer maintained at pH 7.0 and $37°\mathrm{C}$ . Figure 4 shows diagrammatically how the demineralized zone advances in the section overtime. After 6 days, some of the samples cavitated as the simulated lesion advanced completely through the section.

Fig. 3

Automated goniometer with index matching bath to measure angular resolved scattering distributions

Fig. 4

Schematic of the tooth sections of $200\text{-}\mathrm{\mu }\mathrm{m}$ section after 1 and 5 days of pH cycling. The dark zone is sound enamel while the outer zone is the demineralized area.

2.4.

Near-Infrared Attenuation Measurements and Images

The optical imaging system used in this study is a $\mathrm{InGaAs}$ focal plane array ( $318\times 252$ pixels), the Alpha NIR™ (Indigo Systems, Goleta, California) with an infinimite™ lens (Infinity, Boulder, Colombio). Light from a single-mode fiber pigtail coupled to a 1310-nm superluminescent diode (SLD) with an output power of $15\mathrm{mW}$ and a 35-nm bandwidth, model SLED1300D20A (Optospeed, Zurich, Switzerland), was coupled to a 20-mm NIR fiber collimator ( $\mathrm{\mu }\mathrm{LS}$ Micro Laser Systems, Garden Grove, California). Neutral density filters were used to attenuate the incident laser beam. The sample was mounted on an acrylic post and inserted in a cuvette filled with deionized water. The acquired 12-bit digital images were analyzed using IR Vista™software (Indigo Systems, Goleta, California). The setup is shown in Fig. 5 .

Fig. 5

Near-IR imaging setup: (a) $\mathrm{InGaAs}$ FPA, (b) cuvette and section holder, (c) aperture, (d) neutral density filters, and (e) laser fiber pigtail and collimator.

3.

Results

Digital microradiography and near-IR transillumination images of each tooth section with a clearly identified lesion area were acquired. Angularly resolved goniometer measurements were subsequently taken at selected locations in both lesion and sound areas of the tooth section. We present measurements on ten such tooth sections taken from ten natural lesions. The tooth sections were approximately $200\mathrm{\mu }\mathrm{m}$ thick. The microradiographs for the first lesions are shown in Fig. 6 . Figure 6 contains an image of the processed x-ray data that is presented as volume percent mineral—sound enamel is approximately 91% vol.% mineral. The line profiles shown in the bottom of Fig. 6 were taken across the lesion and reflect the severity and topography of the lesion. Goniometer measurements from the tooth section shown in Fig. 6 yielded ${\mu }_s=9$ , $g=0.98$ , and $f=0.25$ for the sound enamel (91% mineral), and ${\mu }_s=190$ , $g=0.97$ , and $f=0.10$ for the most demineralized region (68% mineral). Entire near-IR images of the optical attenuation through the tooth sections were acquired using a large fiber collimator (20-mm beam diameter) and an $\mathrm{InGaAs}$ focal plane array (FPA). Those images are shown in Fig. 7 . Two images are displayed, the top images are the raw intensity images and the lower images represent the attenuation through the sample, which is calculated using the background intensity image, the section thickness, and Beer’s law. The white areas in the processed images represent more opaque regions, and the dark areas are more transparent. The attenuation in the lesion approaches 100 to $200{\mathrm{cm}}^{-1}$ , a factor of 1 to 2 orders of magnitude higher than the surrounding sound enamel, i.e., 2 to $3{\mathrm{cm}}^{-1}$ , and is a similar magnitude to that of the more highly scattering dentin. The absorption coefficient of water at $1310\mathrm{nm}$ is $1.2{\mathrm{cm}}^{-1}$ ,³⁰ and enamel is 8 to 10% water, therefore the absorption coefficient is assumed to be approximately $0.12{\mathrm{cm}}^{-1}$ , a factor of 20 to 30 less than the scattering coefficient, hence we assume that it is negligible.

Fig. 6

(top) A high-resolution digital microradiograph of the lesion area shows the volume percent mineral versus position in the lesion. (bottom) Volume percent mineral versus position for the horizontal line drawn through the lesion.

Fig. 7

(top) Unprocessed images of near-IR $\left(1310\mathrm{nm}\right)$ and (bottom) processed images of NIR optical attenuation through a $160\text{-}\mathrm{\mu }\mathrm{m}$ (left) and $230\text{-}\mathrm{\mu }\mathrm{m}$ (right) thick tooth sections with natural caries lesions are displayed. The attenuation coefficient is in units of ${\mathrm{cm}}^{-1}$ and ranges from 0 to 200.

Angularly resolved scattering measurements were taken from three points of varying mineral content in the lesion of Fig. 8 . One point was the sound enamel (85% mineral), the second point was taken in the less severely demineralized surface zone (78% mineral), and the third point was taken from the most severely demineralized part of the lesion, marked A in Fig. 9 , that is 68% mineral. The ballistic light is reduced by two orders of magnitude and the intensity of the light scattered at angles greater than $10\mathrm{deg}$ is 3 to 4 orders of magnitude higher for the carious tissue over the sound enamel. The angular-resolved goniometer measurements and the corresponding Monte Carlo simulations are shown in Fig. 9. The parameters ${\mu }_s=3.0$ , $g=0.96$ , and $f=0.3$ fit the sound enamel (85% mineral) quite well. In the surface zone marked B (78% mineral) in Fig. 9, the parameters ${\mu }_s=30$ , $g=0.90$ , and $f=0.55$ were used to provide the best fit, while in the most severely demineralized zone in Fig. 9A (68% mineral), the parameters ${\mu }_s=250$ , $g=0.95$ , and $f=0.55$ provided the best fit.

Fig. 8

(top) A high-resolution digital microradiograph of the lesion area shows the volume percent mineral versus position in the lesion area. (bottom) The fraction of NIR $\left(1310\mathrm{nm}\right)$ light scattered versus angle in the sound (blue) and carious (black) areas of the sample measured with the scattering goniometer. Monte Carlo simulations of those goniometer measurements are shown in Fig. 9 (open circle in surface zone is caries in Fig. 9A, filled circle in lesion body is caries in Fig. 9B and cross hatched circle is sound enamel in both Figs 9A, 9B).

Fig. 9

Angular-resolved scattering measurements of sound enamel and carious enamel taken at three positions of varying demineralization in the caries lesion of Fig. 8. Graph A shows the scattering lesion body (filled circle in Fig. 8). Graph B shows scattering from the lesion surface zone (open circle in Fig. 8). The best fit MC simulations are shown as well and the parameters are listed in the text for the MC simulations.

Similar NIR and DM images were acquired from ten caries lesions, and a plot of the optical attenuation versus volume percent mineral loss for ten data points assembled from all the line-profiles across the most highly demineralized area of each lesion are shown in Fig. 10 (note points from each independent sample follow a similar behavior). Nonlinear regression using an exponential growth curve shows that the optical scattering increases exponentially with mineral loss and that the magnitude of light scattering approaches a maximum after 10 to 15% mineral loss, which is consistent with the formation of pores. For higher volume percent mineral loss, additional pores interconnect and the enamel falls apart, forming a cavity above 40 to 50% mineral loss.

Fig. 10

Optical attenuation $\left({\mathrm{cm}}^{-1}\right)$ at 100 points taken from ten natural enamel caries lesions (10 points per lesion). The solid line represents the best exponential fit, $r^2=0.74$ to the attenuation coefficient versus volume percent mineral loss.

3.1.

Artificial Lesions

Angular resolved scattering measurements measured during the development of enamel lesions over a period of five days of pH cycling also show exponential growth. There is a rapid order of magnitude increase in optical scattering after just 1 day of lesion growth, followed by a slower increase of another order of magnitude after the next 4 days.

One of the angularly resolved scattering distributions measured at $1310\mathrm{nm}$ after exposure to five days of a pH cycling regimen is shown in Fig. 11 , along with the same section before lesion formation (sound enamel). The ballistic transmission falls by 2 to 3 orders of magnitude after lesion formation. Although the scattered light appears at larger angles, the scattering distribution is far from isotropic in contrast to our predictions. Apparently, the scattering caused by the pores formed in the enamel as a result of demineralization is highly forward directed. Figure 12 shows the average scattering distributions measured for all five sections after 0 to 5 days of pH cycling. The most dramatic change in the scattering distributions occurred after only one day of pH cycling. Both the dip and rise past $130\mathrm{deg}$ on the scattering distributions is caused by internal reflection of the laser light and refocusing by the scattering chamber, and is not real structure in the scattering distribution.

Fig. 11

Comparison of angularly resolved scattering distributions measured for sound enamel and demineralized enamel produced by exposure to five days of a pH cycling regimen. The MC simulations are shown for sound enamel (day 0) ${\mu }_s=10$ , $g=0.96$ , and $f=0.45$ , and carious enamel (day 5) ${\mu }_s=245$ , $g=0.97$ , and $f=0.05$ provided the best fit.

Fig. 12

Angularly resolved scattering distributions of $200\text{-}\mathrm{\mu }\mathrm{m}$ -thick sections of enamel exposed to pH cycling for a 0 to 5 day section measured at $1310\mathrm{nm}$ . Each trace represents the average of five measurements on five different samples.

The Monte Carlo method was employed to determine the scattering coefficient and the parameters of the scattering phase function that best fit the average curves of Fig. 12, which represent the progressively more demineralized enamel after each day of cycling. Those data are shown in Table 1 . The measured angular resolved scattering distribution used is a linear combination of a highly forward peaked phase function $\Phi \left(\cos \theta \right)$ , a Henyey-Greenstein (HG) function $g$ , and an isotropic phase function $f_d$ represented by Eq. 1. It is important to point out that the samples were immersed in saline and therefore we were unable to eliminate contributions from surface scattering. Surface scattering would be expected to influence all of the optical parameters to some degree.

Table 1

Best fit Monte Carlo simulations for day 0 to 5 of each trace in Fig. 12 using the phase function in Eq. 1.

Day	μs	g	f
0	10	0.96	0.45
1	115	0.97	0.10
2	120	0.97	0.10
3	180	0.97	0.10
4	220	0.97	0.10
5	245	0.97	0.05

4.

Discussion

During the caries process, pores are formed in the lesion due to partial dissolution of the individual mineral crystals. Such small pores act as scattering centers, strongly scattering visible and near-IR light. These measurements indicate that such porosity caused by preferential dissolution in the lesion increases the scattering coefficient by 1 to 2 orders of magnitude higher over the sound tissue. Higher attenuation would be expected for samples that were not immersed in water due to the higher difference in refractive index of air in the porous enamel. The goniometer measurements indicate that a large fraction of the scattered light in the lesion is highly forward directed in contrast to our original hypothesis. This is not entirely unexpected, since demineralization proceeds along the core of the enamel rods, hence the scattering centers may be fairly large in the micron range and highly anisotropic. This also suggests that many of the pores in the lesion are Mie-like scatterers with scattering centers of a similar dimension to the enamel prisms. Such an observation is important for the interpretation of optical coherence tomographic images, since the lesion contrast is due to an increase in backscattered light from the lesion area.

The problem of discriminating between surface scattering and bulk scattering is important for measuring accurate values for the scattering coefficient of sound dental hard tissues. Samples are immersed in an index matching bath of exactly the same refractive index to reduce the effect of surface scattering. In addition, a Beer-Lambert plot of transmission through the tissue versus tissue thickness is used to remove the influence of surface scattering and any other extrinsic factors. Such an approach is not possible, since carious lesions are both highly variable in mineral content and are highly porous. Due to the high porosity, any index matching agent imbides into the lesion, rendering the lesion transparent. Moreover, since the lesion is optically rough and porous, the surface is an intrinsic part of the lesion that cannot be differentiated from the bulk. Therefore, we have taken the approach of measuring the optical attenuation through sections immersed in water to replicate optical conditions appropriate for the oral environment. However, it is important not to forget that the influence of surface scattering has not been decoupled from the scattering in the bulk of the lesion in these measurements. The optical attenuation through the sound regions of the tooth sections varied between 1 to $30{\mathrm{cm}}^{-1}$ due to the influence of surface scattering. The increase in scattering with demineralization was so profound, however, that variations in surface scattering did not prevent the elucidation of the relationship between the volume percent mineral loss and the attenuation coefficient.

The lesions manifest greater contrast between the sound and carious enamel in the near-IR attenuation measurements than in the microradiographs. The image contrast between sound and demineralized enamel in the near-IR due to changes in light scattering is much higher than the image contrast provided by variations in tissue density that occurs at x-ray wavelengths, which provides the image contrast in dental radiography. Therefore, we expect near-IR imaging methods to be more sensitive to early demineralization than x-rays. This is encouraging, since better contrast between sound and demineralized enamel should enable earlier detection of enamel demineralization. The near-IR and the polarized light microscopy images of the lesions are typically more uniform than the microradiographs, which suggests that there is a nonlinear relationship between the mineral loss and the scattering coefficient, namely a large increase in scattering with slight demineralization, followed by a minimal increase in scattering for more severe lesions. The optical property measurements during the longitudinal development of early artificial demineralization suggested a similar rapid exponential increase in optical scattering during initial lesion development, followed by a more gradual increase in scattering as the lesion severity increases.³¹

In summary, we have established a method for quantifying the optical properties of sound and carious dental hard tissues, employing a combination of optical attenuation measurements, angular resolved goniometer measurements, and MC simulations that can be compared with the local mineral loss measured using high resolution digital microradiography. Using these methods, we will be able to sample lesion areas of varying mineral density in the same lesion and produce plots of mineral loss versus optical attenuation, and use such data to establish the relationship between optical scattering and lesion severity.