1. Introduction

The nature of dental decay or dental caries has changed dramatically over the last 50 years with the addition of fluoride to the drinking water and the widespread use of fluoride dentifrices. Today, almost all new decay occurs in the occlusal pits and fissures of the posterior dentition and the interproximal contact sites between teeth. Such decay, particularly in the early stages, is difficult to detect using the dentist’s existing armamentarium of dental x-rays and the dental explorer, (a metal mechanical probe). Therefore, new imaging technologies are needed for the early detection of such lesions. Moreover, the treatment for early dental decay or caries is shifting away from aggressive cavity preparations that attempt to completely remove demineralized tooth structure toward non-surgical or minimally invasive restorative techniques. In non-surgical therapy, a clinician prescribes antibacterial rinses, fluoride treatments, and dietary changes in attempt to naturally remineralize the decay before it becomes irreversible [1]. The success of this type of therapy is contingent on early caries detection and also requires imaging modalities that can safely and accurately monitor the success of such treatment. Conventional x-rays do not precisely measure the lesion depth of early dental decay, and due to ionizing radiation exposure are not indicated for regular monitoring. These constraints and limitations are the impetus for investigating optical imaging systems that could detect early dental decay, while providing the biologically compatible wavelengths that facilitate frequent screening.

Before the advent of x-rays, dentists used light for the detection of caries lesions. In the past 30 years, the development of high intensity fiber-optic illumination sources has resurrected this method for caries detection [2–8]. Fiber-optic transillumination (FOTI) has been shown to be promising for the detection of interproximal lesions. One digital-based system, DIFOTI^™ (Digital Imaging Fiber-Optic Transillumination) from Electro-Optical Sciences, Inc., that utilizes visible light, has recently received FDA approval [9]. However, the strong light scattering of sound dental enamel at visible wavelengths, 400–700 nm inhibits imaging through the tooth. Previous studies have shown that the magnitude of this scattering decreases exponentially at longer wavelengths [10, 13]. The attenuation coefficients of dental enamel measured at 1310 and 1550-nm were 3.1cm-1 and 3.8 cm-1, respectively. The magnitude of scattering at those wavelengths is more than a factor of 30 times lower than in the visible range (Fig. 1) [11]. This translates to a mean free path of 3.2-mm for 1310-nm photons, indicating that enamel is transparent in the near-infrared (NIR). At longer wavelengths past 1550-nm, the attenuation coefficient is not expected to decrease any further due to the increasing absorption coefficient of water, 12% by volume, in dental enamel [12].

In this paper, we demonstrate that near-IR light at 1310-nm is well suited for the detection and imaging of interproximal caries lesions. Simulated lesions, which represent the increased scattering of dental caries, were placed in plano-parallel dental enamel sections of various thickness. These samples were then illuminated with a polarized broadband NIR light source, and images through each section were acquired using an InGaAs focal plane array. The objectives of these measurements were to establish the effective NIR imaging depth through sound tooth structure and to determine the image contrast between simulated caries lesions with sound enamel. Subsequent analysis of the acquired projection images, indicated high contrast between sound enamel and simulated lesions at 1310-nm. In addition, NIR images of extracted teeth containing natural caries lesions were collected to further demonstrate the clinical potential for imaging interproximal lesions, secondary decay around composite restorations, and cracks and defects in the tooth enamel.

 

Fig. 1. The attenuation coefficient of dental enamel (blue)[10, 11] and water (red) [12].

Download Full Size | PDF

2. Materials and methods

2.1 Sample preparation

Thirty plano-parallel sections of enamel of various thickness, 2,3,4,5,6,6.75-mm, were prepared from non-carious human teeth. These sections were stored in a moist environment to preserve tissue hydration with 0.1% thymol added to prevent bacterial growth. Uniform scattering phantoms simulating dental decay were produced midway through each section by drilling 1-mm diameter × 1.2-mm deep cavities in the proximal region of each sample and filling the cavities with hydroxyapatite paste. A thin layer of unfilled composite resin was applied to the outside of the filled cavity to seal the hydroxyapatite within the prepared tooth cavity.

2.2 NIR imaging

The NIR imaging set-up is shown schematically in Fig. 2. Light from either a fiber-optic bundle coupled to a 150-watt halogen lamp, Visar^™ (Den-Mat, Santa Maria, CA) or a 1310-nm superluminescent diode (SLD) with an output power of 3.5 mW and a bandwidth of 25–30 nm, Model QSDM-1300-5 (Qphotonics Inc., Chesapeake, VA) was used as the illumination source.

 

Fig. 2. Setup used for Near-Infrared Transillumination of whole teeth and tooth sections consists of a broadband light source, crossed linear polarizers, a bandpass filter and a NIR InGaAs focal plane array (FPA).

Download Full Size | PDF

We found that the speckle of narrower bandwidth Fabry-perot diode lasers such as a 50-mW 1310-nm source, Model QLD-1300-50 (Qphotonics Inc., Chesapeake, VA) interfered significantly with image resolution and was not useful for this application. Crossed near-IR polarizers, Model K46-252 (Edmund Scientific, Barrington, NJ), were used to remove light that directly illuminated the array without passing through the tooth. Dental enamel is birefringent, therefore, the polarization state of the light passing through the tooth is altered, thus reducing extinction by the orthogonal linear polarizer. An InGaAs focal plane array (318×252 pixels) the Alpha NIR^™ (Indigo Systems, Goleta, Ca) with a Infinimite^™ lens (Infinity, Boulder, Co) was used to acquire all the images. A 50-nm bandpass filter centered at 1310-nm, Model BP-1300-090B (Spectrogon US, Parsippany, NJ), was used to remove all light outside the spectral region of interest. The acquired 12-bit digital images were analyzed using IRVista^™ software (Indigo Systems, Goleta, Ca). The illuminating light intensity, source to sample distance, and the aperture diameter were adjusted for each sample to obtain the maximum contrast between the lesion and the surrounding enamel without saturation of the InGaAs FPA around the lesion area. Although the 3.5-mW SLD source provided similar image quality to the halogen lamp source, all the images shown in this paper were acquired using the fiber-optic illuminator. Due to the natural tooth contours, the sides near the simulated lesions in the tooth sections were masked with putty to esure that light traveled the full width of the sample. This masking is not applicable in a clinical situation and was not necessary to acquire images of whole teeth.

2.3 Visible and x-ray imaging

A tooth section of minimal sample thickness, 3-mm, was chosen for comparison of the NIR transillumination system with conventional visible light FOTI and x-ray transillumination. For visible light transillumination, the same fiber-optic illuminator was used to illuminate the section and a color 1/3” CCD camera with a resolution of 450 lines, Model DFK 5002/N (Imaging Source, Charlotte, NC), equipped with the same Infinimite^™ lens recorded the projection image. The corresponding x-ray image was acquired by placing the section directly on Ultra-Speed^™ D-speed film (Kodak, Rochester, NY) using 75 kVp, 15 mA, and 12 impulses.

2.4 Image analysis

The coordinates of each simulated lesion were known prior to analyzing the contrast of each lesion. The mean pixel intensity of the lesion and the enamel above and below the lesion was measured using the IRvista^™ software. Lesion contrast was calculated for each sample as follows:

where I_E is the mean intensity of the enamel bordering the lesion and I_L is the mean intensity of the lesion. Lesion contrast is defined as a ratio that will vary from 0 to 1. For each of the six sample thicknesses measured, the mean lesion contrast was calculated and plotted versus sample thickness.

Although contrast is important, the boundary or edge between the lesion and the sound tooth structure is central to detection of the lesion. Therefore, the spatial intensity profile of a lesion with its surrounding enamel was analyzed. An intensity profile was mapped from a distinct line in six sample images representing each thickness.

3. Results

Visible light, NIR and x-ray images of a simulated lesion placed in one of the 3-mm thick tooth sections are shown in Fig. 3. The lesion cannot be seen using visible light transillumination, however the lesion is clearly visible with high contrast using NIR light transillumination. A radiographic image of the tooth section using D-speed film shows poor contrast between the lesion and the surrounding enamel.

The lesion contrast was calculated for all thirty of the enamel sections under NIR illumination. Representative spatial intensity profiles from six of the samples of each thickness and the corresponding images are shown in Fig. 4. From these profiles, the edge or boundary between the sound enamel and the lesion is clearly demarcated in all six of the sections. The image contrast plotted vs. section thickness is shown in Fig. 5. A lesion contrast of greater than 0.35 was seen in all the sections with the exception of the 6-mm samples. A 0.35 lesion contrast is equivalent to a lesion intensity that is 65% of the surrounding enamel. For 6-mm samples, a mean lesion contrast of 0.16 was calculated. A steep intensity gradient is visible between the surrounding enamel and the lesion. This gradient is less pronounced for sections greater than 4-mm thick, especially on the lower border of the lesion. A NIR image of a whole tooth sample with a natural lesion, depicted in Fig. 6, illustrates that a natural lesion can be resolved with the same success as the simulated lesions placed in plano-parallel sections. A composite filling is also visible on the opposite side of the tooth in Fig. 6, indicating that there is also high contrast between composite filling materials and sound tooth structure. In the supplemental multimedia file, tooth1, a metal rod was inserted between the illuminating source andthe tooth and rotated approximately 45° to demonstrate that the tooth enamel is transparent. The rod is clearly resolved through the enamel, but the base of the rod that extends beneath the dentin cannot be resolved because of the high opacity of dentin, even in the NIR.

 

Fig. 3. A) Side-view of a 3-mm thick tooth section with a simulated lesion. B) The lesion cannot be seen using transillumination with visible light and a CCD camera. C) The lesion is clearly visible under NIR transillumination D) An x-ray of the section using D-speed film indicates the small contrast difference between the simulated lesion and sound enamel. The sound enamel [e] and dentin [d] layers are distinguishable in all the projection images.

Download Full Size | PDF

4. Discussion

The high contrast and intensity profiles of the simulated lesions with the surrounding enamel indicate the significant potential of NIR transillumination for imaging dental caries. Since the clinical use of transillumination is to detect interproximal lesions, it is important to note that forty of the sixty interproximal surfaces in the mouth would require imaging through less than 5-mm of enamel. This study suggests that resolving caries lesions through 5-mm of enamel is clinically feasible. This is further demonstrated by the NIR imaging of whole teeth with natural decay.

During the transillumination of whole tooth samples, polarization gating with crossed polarizers was critical for preventing the illuminating light from saturating the InGaAs array near the area of the lesion ‘shadow’. This technique will also be important in a clinical setting where adjacent tooth surfaces will reflect, but not depolarize the light, and could interfere with the accuracy of the projection image.

Previous groups pursuing visible light transillumination, have used or proposed more advanced imaging techniques like temporal or coherence gating and sophisticated image processing algorithms to enhance the imaging and detection of dental decay [9, 14]. With NIR transillumination, these techniques may not be necessary.

During demineralization of enamel in the caries process, preferential dissolution of the mineral phase creates pores that highly scatter light. The simulated lesions in our study are primarily made up of isotropic scatterers, with scattering occurring at the grain boundaries in the hydroxyapatite powder [10]. Therefore, such simulated lesions may possibly overestimate the magnitude of scattering in natural caries lesions; however, creating more accurate optical simulated lesions requires an intimate understanding of the fundamental optical properties of carious tissue that has yet to be determined. This study shows that 1310-nm is optimal for both high transmission through sound dental enamel and for achieving high contrast between caries lesions and sound enamel. Simulated lesions composed of an unorganized paste of hydroxyapatite, strongly scatter the 1310-nm light, which provides high contrast with the transparent sound enamel. Optical transillumination is similar to other projection imaging modalities like conventional x-rays, however the image contrast arises from changes in tissue scattering as opposed to variations in tissue density. Therefore, this method can be more sensitive than x-rays for resolving early caries lesions. Clinicians are trained to diagnose at the low lesion contrast depicted in the radiograph of Fig. 3, but the high contrast in the NIR image suggests that the simulated lesions are more sensitive to optical detection. This is due to the fact that the simulated lesions have only slightly lower density than the sound enamel but strongly scatter NIR light. More in depth analysis will be done to investigate the relationship between density loss in caries lesions and their optical properties. A detailed understanding may aid in discriminating dental decay at different stages of the caries and remineralization process. Future studies will investigate and compare the diagnostic ability of NIR transillumination of carious full tooth samples with other detection methods.

 

Fig 4. NIR transillumination images of tooth sections with simulated lesions are shown for each representative sample thickness. The corresponding spatial line profiles demarcated in red are shown on the inset in the lower right of each image, and the measured lesion contrast is shown in the lower left. The left axis represents the pixel intensity ranging from 0 - 4096, and the bottom axis the pixel position through the lesion.

Download Full Size | PDF

 

Fig. 5. The mean ±s.d lesion contrast plotted versus the thickness of the plano-parallel enamel samples, n=5.

Download Full Size | PDF

 

Fig. 6. (2.5 MB) NIR image of a whole tooth sample. A natural carious lesion and a composite restoration are seen on the left and right, respectively. The tooth is slightly rotated to present different viewing angles. A crack is also visible in the center of the tooth.

Download Full Size | PDF