1.

Introduction

Laser-induced fluorescence (LIF) has been increasingly used recently as a powerful tool for caries detection and shows great potential for development as an alternative to the dental probe or radiographic examination. The phenomenon of fluorescence is based on the principle that the fluorescence emission observed on irradiation by light in the blue-green spectral region from carious dental tissues differs from that of a sound tooth. The carious region appears darker than the surrounding sound tooth surface in a fluorescence image due to a decrease in the fluorescence signal from the carious tissues.^{1, 2} A number of optical methods have been developed and recommended as diagnostic tools to identify and quantify early caries lesions on smooth and occlusal surfaces.³

Most of the investigations in caries research have been related with quantitative laser/light-induced fluorescence (QLF), which measures the intrinsic fluorescence of the tooth tissues. Various groups have reported results on the spectral changes in the fluorescence intensity between sound and carious teeth.^{4, 5, 6} The appearance of fluorescence emissions in the $520\text{-}\text{to}580\text{-}\mathrm{nm}$ wavelength region after irradiation with a 488- or $514\text{-}\mathrm{nm}$ argon-ion laser was observed in several studies of carious structures.^{1, 2, 5, 6, 7} Most of these studies showed good correlation between fluorescence and mineral loss^{7, 8, 9} and were found useful in early detection of smooth surface caries lesions.^{7, 10}

Another commercially available laser-based instrument is the DIAGNOdent (KaVo, Biberach, Germany), which uses a $655\text{-}\mathrm{nm}$ diode laser to detect near-IR fluorescence from porphyrins produced by oral bacteria and displays fluorescence of dental tissues as a numerical value ranging from 0 to 99 (higher values for deeper caries). This device has been extensively used in various studies to detect changes in mineral content, to measure changes during remineralization, and in the detection of caries lesions.^{11, 12, 13, 14, 15} However, this device could not provide good correlation with mineral loss on smooth surface caries lesions in teeth.^{9, 16, 17} Moreover, despite the fact that some authors have claimed that the device is able to detect early enamel caries lesions,^{18, 19} studies have not yet been conclusive.^{17, 20}

The most preferred excitation wavelength for caries detection by LIF spectroscopy is around $400\mathrm{nm}$ as this wavelength provides optimal excitation of tissue fluorophors.^{21, 22, 23} Thus, a diode laser with emission around $400\mathrm{nm}$ is the preferred choice as compared to the broad emission bands produced by nonlaser excitation sources.

The aim of this study was to develop a LIF ratio reference standard (FRRS) to discriminate between different stages of caries lesions using a diode laser emitting at $404\mathrm{nm}$ as the excitation source. Toward this, we measured the in vivo LIF spectra from sound and carious teeth of 65 patients and developed fluorescence intensity (FI) and FRRS scatter plots based on autofluorescence spectral intensity of the emission peaks at 500, 635, and $680\mathrm{nm}$ and FI ratios $F500∕F635$ and $F500∕F680$ , respectively. The sensitivity and specificity of the FI at the emission peaks and ratios in caries discrimination was determined after validation by clinical examination of caries lesions and the results are presented.

2.

Material and Methods

2.3.

Instrumentation

The portable laser-induced fluorescence spectroscopy (LIFS) system developed in our laboratory for caries diagnosis is shown in Fig. 1 . It is composed of a diode laser (Stocker Yale, Canada, $404\mathrm{nm}$ , $50\mathrm{mW}$ , cw) for excitation of tooth fluorescence and a miniature fiber optic spectrometer (Ocean Optics, United States, Model: USB 2000FL VIS-NIR), connected to the USB (universal serial bus) port of a computer for spectral recording. A bifurcated optical fiber guides the light from the laser to the tooth through a $3\text{-}\mathrm{m}$ -long optical fiber ( $400\mu \mathrm{m}$ diameter) terminated in a hand piece made of stainless steel. A $5\text{-}\mathrm{mm}$ -long black PVC sleeve inserted at the probe tip helps to prevent stray room light from entering into the detection system by providing a close contact with the tooth surface. Since this opaque sleeve is disposable, it provides extra hygiene. The light emanating from the sample is collected by the same optical fiber and delivered to a miniature fiber optic spectrometer through the second arm of the bifurcated fiber and the long-wavelength pass filter (Schott: GG420).

Fig. 1

Schematic of the experimental setup for LIF spectral measurements.

3.

Results

3.1.

LIF Spectral Features

In vivo fluorescence spectral measurements were carried out on consenting patients having caries lesions in their oral cavities. Figure 2 shows the mean LIF spectra recorded from 25 sites of sound teeth, 30 sites of enamel caries, and 35 dentinal caries lesions, and the same spectra normalized with respect to the maximum spectral intensity around $500\mathrm{nm}$ are given in Fig. 2. The standard deviation is shown for the prominent peaks in the mean LIF spectra from sound and caries teeth belonging to different stages of tooth decay. Marked differences in fluorescence spectral features were seen between sound and caries teeth. As compared to sound teeth, the overall FI was found to be lower in all caries lesions. However, the fluorescence spectral intensity in the red wavelength region was greater for caries lesions than for sound teeth. The LIF spectrum of sound teeth shows a broad emission around $500\mathrm{nm}$ with a long tail extending toward the red wavelength region. In enamel and dentinal caries, two additional peaks were seen around 635 and $680\mathrm{nm}$ owing to porphyrin emission. These porphyrin peaks are more prominent in dentinal caries lesions. Another notable feature is the broadening of the $500\text{-}\mathrm{nm}$ peak towards the red wavelength region by about $30\mathrm{nm}$ in dentinal caries. The spectra also shows a sharp rising edge in the short-wavelength side $\left(400\text{to}450\mathrm{nm}\right)$ , which is due to absorbance of the $420\text{-}\mathrm{nm}$ long-wavelength pass filter (Schott GG420) used for blocking the back-scattered laser light from entering the spectrometer.

Fig. 2

(a) Mean in vivo LIF spectra of sound teeth and those of different caries stages and (b) LIF spectra normalized to maximum fluorescence spectral intensity around $500\mathrm{nm}$ . Sound tooth spectra represent the mean of 15 measurements each in 25 sites, whereas enamel caries spectra are of $15\times 30$ measurements and the dentinal caries spectra are of $15\times 35$ measurements.

Figure 3 shows the mean LIF spectral intensity variation of emission peaks at 500, 635, and $680\mathrm{nm}$ from sound teeth, enamel, and dentinal caries. The LIF intensity at $500\mathrm{nm}$ decreases as caries progression occurs in the sound tooth, whereas the intensity increases for the porphyrin peaks at 635 and $680\mathrm{nm}$ . A one-way ANOVA test was used to determine the average spectral intensity variation among the three groups studied. Statistically significant differences in mean spectral intensities $(p<0.001)$ were noticed between sound teeth, enamel caries, and dentinal caries lesions.

Fig. 3

Mean LIF spectral intensity variation of emission peaks at 500, 635, and $680\mathrm{nm}$ from 65 patients diagnosed as enamel caries and dentinal caries, along with the intensity values from 25 sites of sound teeth.

3.3.

Caries Discrimination

3.3.1.

Using FI scatter plots

To discriminate different stages of caries lesions from sound teeth, scatter plots were drawn based on the fluorescence intensities of emission peaks at 500, 635, and $680\mathrm{nm}$ obtained from the spectral data of 65 carious sites and 25 sound teeth sites [Figs. 4, 4, 4 ]. The cutoff value in the scatter plot, which is the weighted mean of the adjacent groups, is used to discriminate different stages of caries lesions. For example, the cutoff values for differentiating sound teeth from enamel caries, sound teeth from dentinal caries, and enamel and dentinal caries are 843.9, 662.4, and 483.0, respectively, at $500\mathrm{nm}$ . Discrimination lines were drawn between sound teeth and enamel caries, sound teeth and dentinal caries, and enamel and dentinal caries in the three scatter plot diagrams. Diagnostic accuracies, such as sensitivity, specificity, positive predictive value, and negative predictive value, for discriminating each pair were calculated by correlating the position of emission intensity values for each lesion in the scatter plot with the corresponding clinical criteria.

Fig. 4

FI scatter plots based on autofluorescence spectral intensity of emission peaks, at (a) 500, (b) 635, and (c) $680\mathrm{nm}$ , from 65 patients diagnosed as enamel caries and dentinal caries, along with the emission intensities from 25 sites of sound teeth.

Table 3 illustrates the diagnostic accuracies obtained for classifying different stages of caries lesions using fluorescence emission intensities at 500, 635, and $680\mathrm{nm}$ . The sensitivities obtained for discriminating sound versus enamel caries, sound versus dentinal caries, and enamel versus dentinal caries pairs are 83, 97, and 86%, respectively, and the corresponding specificities are 88, 88, and 73%, respectively, using the fluorescence emission intensity at $500\mathrm{nm}$ . For 635- and $680\text{-}\mathrm{nm}$ emission peaks, the sensitivities and specificities were very poor for all the lesion pairs, except for the moderate specificities of 92 and 96% observed respectively for these two peak intensities in discriminating sound teeth from dentinal caries.

Table 3

Diagnostic accuracies obtained for classifying different stages of dental caries using the fluorescence intensity of emission peaks at 500, 635, and 680nm .

FI	Sound vs Enamel Caries (%)				Sound vs Dentinal Caries (%)				Enamel vs Dentinal Caries (%)
	Se	Sp	PPV	NPV	Se	Sp	PPV	NPV	Se	Sp	PPV	NPV
500	83	88	89	81	97	88	92	96	86	73	79	81
635	50	64	63	52	31	92	85	49	31	73	58	48
680	50	68	65	53	43	96	85	55	43	83	75	56

FI, fluorescence intensity; sensitivity (Se), true positive/(true positive+false negative); specificity (Sp), true negative/(true negative+false positive); positive predictive value (PPV), true positive/(true positive+false positive); and negative predictive value (NPV), true negative/(true negative+false negative).

3.3.2.

Using FRRS scatter plots

Figures 5 and 5 illustrate the scatter plots of the fluorescence intensity ratios $F500∕F635$ and $F500∕F680$ in 65 patients, with tooth lesions diagnosed as enamel and dentinal caries, along with the control data from 25 sites of sound teeth. The intensity ratios of caries lesions are generally lower than those of sound teeth (Table 2) and have low independent Student’s $t$ test values of $p<0.001$ . Discrimination lines were drawn between sound teeth and enamel caries, sound teeth and dentinal caries, and enamel and dentinal caries as in the case of FI for the two scatter plot diagrams. Diagnostic accuracies for discriminating each pair were calculated by correlating the position of ratio values for each lesion in the scatter plot with the corresponding clinical criteria.

Fig. 5

FRRS scatter plots based on autofluorescence spectral intensity ratios (a) $F500∕F635$ and (b) $F500∕F680$ from 65 patients diagnosed as enamel caries and dentinal caries, along with the ratios from 25 sites of sound teeth.

Table 4 illustrates independent and overall sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of FRRS with respect to $F500∕F635$ and $F500∕F680$ for classifying different stages of tooth caries. For the $F500∕F635$ and $F500∕F680$ ratios, by selecting a cutoff at the mean of sound tooth and dentinal caries values (6.87 and 16.67) respectively, sensitivities and specificities of 100% were achieved with a positive and negative predictive value of 1 for discriminating these two categories. In Fig. 5, by selecting 8.69 as the cutoff value in the $F500∕F635$ scatter plot to discriminate sound teeth from enamel caries, sensitivities and specificities of 87 and 100%, respectively, were obtained with a PPV of 1 and an NPV of 0.86. In this study, only 4 out of the 30 enamel caries lesions were misclassified as sound. For discriminating enamel from dentinal caries, sensitivities and specificities of 89 and 80%, respectively, were achieved for the same ratio by using 4.73 as the cutoff value, with a corresponding PPV of 0.84 and an NPV of 0.86. Here, only 4 out of 35 dentinal caries lesions were misclassified as enamel and 6 out of 30 enamel caries lesions were misdiagnosed as dentinal caries. Furthermore, overall sensitivities and specificities of 85 and 90%, respectively, were achieved for discriminating sound teeth from enamel caries, whereas sensitivities and specificities of 100% were obtained to discriminate sound teeth from dentinal caries with both the FRRS scatter plots.

Table 4

Independent and overall diagnostic accuracies of FRS ratios F500∕F635 and F500∕F680 obtained for classifying different stages of dental caries.

LIF ratios	Sound vs Enamel Caries (%)				Sound vs Dentinal Caries (%)				Enamel vs Dentinal Caries (%)
	Se	Sp	PPV	NPV	Se	Sp	PPV	NPV	Se	Sp	PPV	NPV
$F500∕F635$	87	100	1.00	0.86	100	100	1.00	1.00	89	80	0.84	0.86
$F500∕F680$	83	80	0.83	0.80	100	100	1.00	1.00	86	73	0.79	0.82
Overall	85	90	0.92	0.83	100	100	1.00	1.00	88	77	0.82	0.84

Sensitivity (Se), true positive/(true positive+false negative); specificity (Sp), true negative/(true negative+false positive); positive predictive value (PPV), true positive/(true positive+false positive); and negative predictive value (NPV), true negative/(true negative+false negative).

4.

Discussion

Currently visual inspection or its combination with radiography is considered as the gold standard for caries detection. However, x rays are ionizing and hazardous in nature and cannot detect caries until they are well advanced. Hence, the development of diagnostic algorithms based on LIF spectral ratios would be of great help for the real-time, noninvasive detection of dental caries in a clinical environment.

In this clinical trial, we observed significant differences in the autofluorescence spectra of sound and carious teeth. This disparity in fluorescence spectra is based on the absorption and scattering properties of light by cariogenic substances and fluorophores in the tooth. The normal tooth enamel is composed of millions of prisms or rods with waveguide properties that facilitate deep penetration when illuminated with visible light. In the case of dental caries, the prism structure is damaged and the waveguide properties are lost so that the irradiated light cannot penetrate deeply as with sound enamel. This leads to a reduction in the fluorescence intensity in caries lesion.⁵

Another reason for fluorescence spectral variations might be due to the import of exogenous fluorescent molecules during the caries process. This fact is clearly supported by the progressive rise of fluorescence spectral intensity in the red wavelength region with caries progression, and the consequent decrease in the autofluorescence emission around $500\mathrm{nm}$ (Fig. 2).

The autofluorescence emission spectrum of tissue is basically a convolution of the emission spectra of the endogenous fluorophores of tissue and therefore strongly depends on the wavelength of excitation light.²⁵ Only those endogenous fluorophores whose absorption bands have an overlap with the wavelength of excitation light are excited and emit fluorescence. Since the excitation and emission light must propagate through the turbid tissue, the recorded autofluorescence is also influenced by the absorption and scattering at both the excitation and the emission wavelengths. These endogenous fluorophores include the aminoacids, structural proteins such as collagen and elastin, coenzymes such as NADH and flavins, vitamins, lipids, and the porphyrins. Their excitation maxima lie between 250 and $450\mathrm{nm}$ , whereas their emission maxima lie²⁶ in the range of 280 and $700\mathrm{nm}$ .

Fluorophores that are speculated to play a major role in the caries processes are structural proteins such as collagen, porphyrins, and bacteria. Collagen and elastin are fibrous proteins and are abundant in connective tissues, teeth, and bones. Collagen forms the organic part of the dentin and any structural or pathologic association with caries processes could be reflected in lower autofluorescence intensity.²⁷ Therefore, these factors could be responsible for the similarity observed between the reference standards for classification of tooth caries and oral cavity cancer from tissue autofluorescence.²⁸

The broad autofluorescence around $500\mathrm{nm}$ is due to emission from natural enamel²² and peaks observed at 635 and $680\mathrm{nm}$ are due to emission from endogenous porphyrins and metalloporhyrins, in particular, protoporphyrin IX (PpIX), mesoporphyrin, and coproporphyrin synthesised by bacteria.^{18, 22} PpIX is reported to be higher in gram-negative oral bacteria and its concentration increases as the dental biofilm becomes more mature, which is responsible for the red fluorescence in teeth.²⁹ Likewise, Buchalla²¹ reported that caries lesions fluoresce at 624, 650, and $690\mathrm{nm}$ due to the presence of porphyrins, and the fluorescence efficiency increases when excitation is in the wavelength range between 400 and $420\mathrm{nm}$ .

Significant differences in the FI ratios ( $F500∕F635$ and $F500∕F680$ ) were observed during the caries process. The mean fluorescence spectral intensity ratios of sound teeth (Table 2) were always found to be higher as compared to carious teeth. The decrease in the FI ratio values point to the extent of teeth decay. Both these ratios that are used to classify the sound tooth from caries have very low independent Student’s $t$ test values, $p<0.001$ . Moreover, using the diagnostic algorithm developed based on the FRRS scatter plots, we obtained overall sensitivities of 85, 100, and 88%, respectively, for discriminating sound teeth from enamel caries, sound teeth from dentinal caries, and enamel caries from dentinal caries with corresponding specificities of 90, 100, and 77% (Table 4). In comparison, the FI scatter plots of fluorescence emission intensity at $500\mathrm{nm}$ gave sensitivities of only 83, 97, and 86%, respectively, for discriminating sound teeth from enamel caries, sound teeth from dentinal caries, and enamel from dentinal caries with corresponding specificities of 88, 88, and 73% (Table 3). The advantage of using scatter plot ratios over peak intensity is that it helps to eliminate errors associated with changes in excitation energy or incident light levels, collection efficiency, and the spectral response of the detection system.

In comparison, Pinelli, ¹⁴ who reported a sensitivity of 72% and a specificity of 73% for detecting active and arrested incipient caries lesions on smooth surfaces using DIAGNOdent, proposed that it could be used as a good auxiliary method along with clinical examination. In another study, Shi ⁹ reported a sensitivity of 94% and a specificity of 100% for detecting smooth surface caries with the QLF method, but achieved a lower sensitivity and specificity of 75 and 96%, respectively, using DIAGNOdent. Similarly, Lussi ¹⁵ reported a sensitivity and specificity of 92 and 86%, respectively, for detecting occlusal dentin caries using the DIAGNOdent device as compared to visual inspection and bitewing radiography. Furthermore, laser-based techniques are reported to be effective in caries diagnosis, as they combine the advantages of higher specificity and speed of visual inspection with the higher sensitivity.¹⁵ Also, Burin ³⁰ assessed the efficiency of LIF, visual examination, and bitewing radiography and found that visual inspection was as valid as LIF, which should be considered a better adjunct than bitewing radiography for caries diagnosis.

This investigation shows that FRRS algorithms based on the autofluorescence intensity ratio of the emission peaks can localize and discriminate different stages of caries lesions. However, for a more reliable caries diagnosis, the combined use of visual inspection and LIFS is recommended, since the laser has the benefit of quantifying mineral content and helping dentists to improve diagnostic efficacy and treatment.

5.

Conclusion

The results of this clinical study illustrate that information provided by noninvasive LIFS has excellent potential to discriminate different stages of dental caries. The FRRS diagnostic algorithm based on tissue autofluorescence spectral ratio was able to discriminate sound teeth from enamel caries, sound teeth from dentinal caries, and enamel from dentinal caries with sensitivities of 85, 100, and 88%, respectively, with ensuing specificities of 90, 100, and 77%. The study proves that LIFS could function as an auxiliary method to dentists for detecting and discriminating dental caries in a fast and sensitive manner. Our results confirm that classification of tooth caries from fluorescence signatures with $404\text{-}\mathrm{nm}$ diode laser excitation enables precise visualization and quantification of both the intrinsic green fluorescence of dental hard tissues and the red fluorescence of bacterial origin.