Open Access
Add to BibSonomy
Abstract
One of the major oral health conditions worldwide is dental caries. Light-absorption-based thermophotonic diagnostic imaging is well positioned for this challenge thanks to its speed, safety, and high molecular contrast advantages. In this work, a multispectral (MS) truncated-correlation photothermal coherence tomography (TC-PCT) imaging modality is introduced for the detection of bacterial-induced dental caries. MS TC-PCT provided thorough information about optimal lesion contrast and type of dental defects such as caries in teeth. The experimental results were validated using micro-computed tomography (µCT) including quantitative lesion depth profiles at wavelengths in the 675-700 nm range. MS TC-PCT gives rise to hard-tissue biomedical diagnostic applications such as bone and dental imaging.
© 2022 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement
1. Introduction
Worldwide, the spectrum of oral health diseases includes dental caries, periodontal diseases, oral cancers, oral infectious diseases, and orodental trauma. Despite their preventable nature, oral diseases affect close to 3.5 billion people worldwide, with 2.3 billion adults suffering from caries of permanent teeth while more than 530 million children suffer from caries of primary teeth [1]. Dental caries arises when plaque forms on the surface of a tooth and converts the sugars contained in our daily meals into acids that destroy the tooth over time [2]. Several assessment modalities have been proposed for dental caries detection. Among them, X-rays are a well-established technique for visualizing the tooth structure, yet they are not sensitive enough to reliably detect tooth demineralization at its early onset and are also unsuitable for routine monitoring due to harmful ionizing radiation [3]. In recent years, optical imaging techniques have exhibited advantages in biomedical imaging due to their non-ionizing radiation, high-contrast, and high-resolution imaging. Among them, while not yet adopted for clinical use in dentistry, optical coherence tomography (OCT) is a non-invasive, non-ionizing diagnostic imaging tool based on optical interferometry. OCT achieves an excellent spatial resolution by using a low-coherence broadband near-infrared light source. OCT, however, exhibits limits in terms of scanning range and penetration depth [4]. Besides, wavelength choice is an important factor affecting the performance of OCT in dental applications. For example, an OCT system with 1550-nm center wavelength is good for hard tissue measurement but not suitable for soft tissue imaging [5]. When it comes to other biophotonic imaging methods, the concept of transillumination for diagnosis is based on the principle that different tissues vary in their refractive index when light is passed through them, based on their composition. Fiber-Optic Trans-Illumination (FOTI) comprises compact devices with a fiber optic tip that should be positioned perpendicular to the facial surfaces illuminating the tooth [6]. With the demineralized region having a lower refractive index than a healthy tooth, its illumination is less, and it appears dark. However, due to overlapping signals from the healthy and carious regions, the specificity and contrast levels of FOTI images are lower. Vaarkamp et al. [7] approached FOTI by using wavelength-dependent light propagation through carious regions and concluded that the method is feasible for quantitative diagnosis of approximal caries lesions and can provide insight into the depth of the lesion. The method, however, it is not applicable in all locations of carious lesions. Another technique is the use of near-infrared wavelengths between 780-1550 nm for trans-illumination of dental enamel. This imaging technique is based on the transmission of light through the enamel tissue to differentiate between intact and carious enamel tissue, however, it is not able to assess the extent of lesions, especially monitoring the progression of early caries [8].
Alternative to these optical based technologies, thermophotonic imaging is an imaging modality that takes advantage of optical-to-thermal energy conversion leading to the detection of IR photons from samples captured by a mid-IR camera in order to provide penetration depth beyond the optical diffusion length. However, in accordance to the Fickian principle, thermal waves follow a parabolic heat diffusion equation and exhibit depth-integrated energy distribution and poor axial resolution for thermophotonic technologies. To overcome the poor axial resolution due to the parabolic (diffusive) nature of thermal waves, a thermophotonic imaging technique has been developed known as enhanced TC-PCT (eTC-PCT) [9]. eTC-PCT uses a chirped pulse excitation with an 808-nm modulated continuous wave (CW) diode laser and cross-correlates the thermal transient signals captured by an IR camera with the in-phase and quadrature reference signals, as can be seen in Fig. 1(a). The resulting cross-correlation is then truncated based on a time gating filter which is calculated based on a user-defined slice width, WT, and millisecond incrementing delay, d. eTC-PCT software reconstructs the depth distribution of photothermal parameters from its main output channels, i.e. amplitude and phase. Use of eTC-PCT and its original version, TC-PCT [10], has grown in a variety of applications from imaging subsurface defects and discontinuities in solids and defectoscopy in art objects [11] to medical applications such as imaging of bone [12] and tumor progression in mice [13]. eTC-PCT harnesses photothermal phenomena to meet the demands for high-resolution 3D imaging of early-phase biomedical and dental lesions. In eTC-PCT, the 808 nm laser diode is used for its waveform convenience, high peak power availability, and highly optically penetrating wavelength in tissue. However, in the conventional eTC-PCT, the laser excitation pulse width of commercial 808-nm diode lasers cannot decrease below the millisecond range, nor can the optical penetration depth change to accommodate very near surface and very deep absorption features.
Fig. 1. (a) eTC-PCT signal processing methodology (b) experimental configuration of multispectral TC-PCT system and photograph of the tooth sample with early caries.
In this work, a new approach has been introduced by incorporating a powerful optical source such as a pulsed Nd:YAG laser which can generate ultra-short (nanosecond) pulses and an optical parametric oscillator (OPO) to convert the laser emission into a multiwavelength signal thereby introducing and developing MS TC-PCT imaging. This enabled the 680-1000 nm wavelength range with maximum energy per pulse of 60 mJ occurring at 700 nm. With MS TC-PCT, better image quality can be attained by increasing the laser pulse energy delivered to the sample at short excitation pulse widths while keeping the energy density within the maximum permissible exposure limit [14]. Ultra-short pulses are capable of maintaining a flat power spectrum over a very broad bandwidth with no or minimal loss, despite diffusive attenuation losses. The higher energy deposition due to longer excitation pulse width such as in conventional eTC-PCT can reduce the depth-resolved properties and spatial resolution of the eTC-PCT reconstruction images. In addition, light propagation in biological tissues, is highly random due to profuse scattering. However, the amount of scattering is a function of the photon energy, that is, OPO wavelengths (lower energy photons) longer than 750 nm are less scattered compared to shorter wavelengths (higher energy photons) between 675-750 nm and can therefore penetrate deeper into the substructure. Although the amount of light scattering is a key factor in determining the light penetration depth, a photothermal signal is generated only when heat is released in an absorption event followed by non-radiative energy conversion. To date, the image reconstruction principle of eTC-PCT on teeth has only been tested on a single wavelength at 808 nm with a CW laser in 40-80 ms of excitation time. The absorption and scattering of light in teeth is dependent on the wavelength of the excitation light with enamel exhibiting various absorption peaks at different wavelengths. Therefore, MS TC-PCT allows the selection of excitation and detection wavelengths for optimal imaging contrast based on both the scattering and absorption of light photons.
2. Methods
A Q-switched Nd:YAG laser (Surelite OPO Plus SLIII-10, Continuum, San Jose, United States) generates 5-ns laser pulses at 10-Hz repetition rate which is the frequency of its flashlamp discharge. With this Q-switched laser, experiments were performed at a single pulse repetition rate and the repetition frequency was calculated using Eq. (1):
where Fxx is the frequency of flashlamp discharge. Pxx is the laser pulse repetition frequency with every discharge of flashlamp and could be changed from 0 to 99.This feature could change the frequency of laser pulse output without changing the flashlamp discharge rate. The laser pulse could be frequency doubled to 532 nm, which then pumped the OPO. The OPO output wavelength is tunable from 675 nm to 1000 nm and is controlled by a designated software to give full control over the crystal motor to achieve the desired wavelength. The wavelength tunable ultra-short laser pulses are the most unique feature of MS TC-PCT compared to an original embodiment of eTC-PCT which uses a ms pulse duration range from an 808-nm diode laser (Jenoptic JOLD- 120-QPXF-2P) as excitation source in a material medium. The experimental setup (Fig. 1(b)) includes an IR camera (A6700sc, FLIR, USA, 3–5 µm spectral response) which records the thermal evolution of the target sample following exposure to laser irradiation. The electrical nanosecond pulse coming from the laser acted as an external trigger and was sent to a function generator (Agilent 33220A, USA), which captured the nanosecond pulse and converted it to a millisecond pulse so that it could be captured by a high-speed data acquisition module (NI PCI-6281) for synthesizing the reference signal. The captured reference signal was then synchronized with the thermal transient signal recorded by the camera. The laser beam was passed through a beam diffuser (ED1-C20, Thorlabs Inc., New Jersey, USA) to become homogenized. Details regarding the signal generation and reconstruction algorithm using eTC-PCT technique can be found elsewhere [9] [15], with the exception that a single repetition frequency was used in the MS TC-PCT algorithm instead of linear frequency modulation (LFM) chirp.
3. Results and discussion
An extracted healthy tooth was used in this study, as can be seen in Fig. 1(b). To closely mimic the natural formation of dental caries, a treatment protocol using highly cariogenic bacteria was used to produce a carious lesion [16]. The bacterial-induced caries was limited to a rectangular section on the smooth surface of the tooth for a total exposure time of 8 days to induce early caries. As a result, the overall appearance of the extracted tooth in visible light remained unchanged. The structure of this sample, as seen in its visible-light image provided an opportunity for a clear demonstration of the tomographic nature of MS TC-PCT data. For MS TC-PCT imaging, key tunable parameters are the wavelength and laser pulse repetition frequency range while the laser excitation pulse width and reference sliced width (WT) stay constant at five nanoseconds and 100 ms, respectively. Both wavelength and frequency range have considerable effect on the reconstructions. Thermal waves at low frequencies can propagate deeper into the sample, while thermal waves at high frequency are limited to near-surface depths. Penetration depth describes the depth of light penetration into a medium when the intensity of the radiation inside the medium falls to 1⁄e of its original value. The wavelength dependence of the scattering cross-section is such that shorter wavelengths (or higher optical frequency) are scattered more strongly than longer wavelengths in dental tissues. Figures 2(a) and 2(b) show the MS TC-PCT amplitude reconstructions of the investigated tooth at single repetition frequencies of 0.2 Hz and 0.5 Hz, respectively, and at OPO wavelengths 532 nm, 675 nm, 700 nm, 750 nm, 808 nm, 850 nm, and 900 nm.
Fig. 2. 3D eTC-PCT amplitude reconstruction of a carious lesion in a tooth sample taken at (a) 0.2 and (b) 0.5 Hz at pulsed Nd:YAG laser OPO wavelengths 532 nm, 675 nm, 700 nm, 750 nm, 808 nm, 850 nm, and 900 nm. Below each 3D reconstruction, the top view of the cross-section of the reconstructed 3D model is shown, cut along the dashed black line, as marked in the 532-nm 3D reconstruction. The extent of the lesion inside the tooth is seen in the top view of the cross-section. Note that the x- and z-axis scales are different in the 3D reconstruction and the top view for better visualization. The bacterial-induced caries is shown by a black arrow in each reconstruction. Red and blue arrows identify a shallow and a deep natural lesion on the tooth, respectively.
In addition, the top view of the transverse cross-section of the reconstructed 3D model, cut along the black dashed line, is shown below each 3D reconstruction for all wavelengths. For 3D reconstruction, the pixel-by-pixel data from the resulting cross-correlation was processed with CADIPT Lab’s customized LabVIEW TC-PCT application software. These data were either directly used for 3D reconstruction through ImageJ software [17] or imported into MATLAB 2021a for data analysis (e.g. normalization and SNR calculation) of the images obtained from the TC-PCT amplitude and phase channels. Using ImageJ for each channel, the slice images are imported as a stack and compiled in 3D using the Volume Viewer plugin in ImageJ. For 3D reconstructions, nearest neighbour interpolation with 1.0 sampling rate was used. To compare the effect of the laser repetition frequency and wavelength in the 3D reconstruction images, it can be seen that the 0.5 Hz Fig. 2(b) produces images with better contrast and SNR compared to the measurements taken at 0.2 Hz, Fig. 2(a). For this application, the SNR values were obtained by finding the ratio between normalized image pixel value and standard deviation of intact reference area. For any given amplitude image, the normalized amplitude pixel value was obtained using Eq. (2).
Table 1. SNR values obtained for amplitude images at different OPO wavelengths
The MS TC-PCT amplitude images show the bacterial-induced caries better at shorter OPO output wavelengths such as 675 nm and 700 nm compared to OPO wavelengths that are longer than 700 nm (the bacterial-induced caries is shown by black arrows). A reason for such a phenomenon is that the caries is at its earliest stages, hence it is a near-surface feature; higher repetition frequencies and shorter wavelengths tend to capture such shallow features due to the shorter thermal diffusion lengths involved and the highly scattering nature of the lesions which results in near-surface photon localization. Regarding the 532-nm images, there is more noise in the dental caries region and this can be due to the highly scattered incident photons at that wavelength which tend to also highlight other surface defects and built-in inhomogeneities. Caries in its early stages usually has an intact and a well-mineralized surface layer covering an area of subsurface demineralization [18]. In amplitude images at high frequencies and/or short wavelengths, this intact layer may dominate the light scattering process while the enhanced absorption due to the demineralization below the intact layer could be masked. At 850 nm and 900 nm, the carious lesion starts to disappear for two reasons: longer wavelengths penetrate deeper into the sample and capture deeper features, thereby decreasing the statistical weight of the carious layer in the image generation process. Also, the energy per pulse of OPO wavelengths declines as wavelength increases, and that contributes to noise. The location and extent of the lesion in the tooth is seen in the top cross-sectional views, with the 0.5-Hz repetition frequency at 675-700 nm offering the optimal visualization of the subsurface extent of the caries. Aside from the bacterial-induced caries, one can see another carious region highlighted by the red arrow that is only captured by the 532-nm wavelength. We conclude that this carious region is positioned very close to the surface that only one highly scattered wavelength, i.e. 532 nm, is able to capture this feature. Early caries also appears at a higher contrast at 0.5 Hz, characterizing and visualizing the close-to-the-surface extent of this early lesion.
Figures 3(a) and 3(b) show the phase images corresponding to Fig. 2. 3D Phase images are emissivity-normalized and are best used in terms of depth probing capability as they probe deeper than amplitude images, thereby detecting lesions at deeper subsurface distances within the enamel. It can be seen that features such as the borders of the carious and healthy areas of the tooth are much less pronounced and less sensitive to noise compared to the amplitude channel, which is due to the phase channel containing information from deeper regions than superficial defects. In the phase images, MS TC-PCT shows higher contrast and detail at 0.5 Hz compared to 0.2 Hz. The phase images at 675 nm and 700 nm show a highly detailed reconstruction of the anatomy of the bacterial-induced caries. Even though the bacterial-induced caries is visible at all phase images except at 900 nm, the top cross-sectional view of the reconstructed 3D model at 0.5 Hz and OPO wavelength of 700 nm is optimal, with the extent of the lesion distinctly visible. Conversely, the feature pointed to by a blue arrow in both amplitude and phase images at 750 nm, 808 nm, 850 nm, and 900 nm is a deeper lesion that is captured only by these OPO wavelengths.
Micro-Computed Tomography (µCT) was conducted to confirm the presence and extent of the demineralized region and other features of the tooth. The presence and location of bacterial-induced caries can be seen in Fig. 4(a) inside the green rectangular region. The actual depth of the carious lesion was verified with µCT and was found to be approximately 300 µm. The red rectangular area in Fig. 4(a) shows the dental caries on the extreme left side of the tooth (the carious lesion is shown by red arrow in all the images). The depth of this caries is approximately 200 µm, and the intact surface layer covering the subsurface demineralization can be clearly seen, thereby validating the surface scattering origin of the poor contrast observed in the 532-nm amplitude images, Fig. 2. The µCT slice (Fig. 4(b)) of the deep lesion that is pointed to by the blue arrow in both amplitude and phase images shows it is indeed a deeper lesion compared to other defects on the tooth with an approximate depth of 400 µm.
Fig. 4. Micro-CT slices of the tooth depicting (a) the shallow caries on the extreme left side of the tooth is indicated by the red arrow inside the red box and the bacterial-induced caries is indicated by the green arrow inside the green box (in Figs. 2 and 3 these two features are shown by red and black arrows, respectively); and (b) a deep natural feature in the tooth shown with blue arrows (blue arrows in Figs. 2 and 3 as well) inside the blue rectangular borders.
To compare the extent of caries found using µCT and MS TC-PCT, the top view of the transverse cross-section in amplitude images of Fig. 2(b) at 675-700 nm OPO wavelength range, shows the extent of the bacterial-induced caries to be approximately 300 µm, which is close to the value obtained from the µCT image. The top view of the transverse cross-section of phase images of Fig. 3(b), shows that at 700 nm, the depth of the bacterial-induced caries is closest to that of the value obtained in µCT. Nevertheless, an extensive analysis between depth values obtained in MS TC-PCT is still in progress toward full quantitative image reconstruction.
4. Summary
The present work explored the development of the MS TC-PCT modality and its application to dental imaging. OPO wavelengths between 675 and 700 nm captured the bacterial-induced caries with high SNR and contrast while the deeper natural lesion identified with the blue arrows in Figs. 2 and 3 was detected at OPO wavelengths of 750-900 nm. The natural caries shown with the red arrow was best detected at 532-nm wavelength due to its shallow depth compared to other lesions in the tooth. The experimental results, specifically the extent of the caries observed in the MS TC-PCT images, were validated using µCT.
In MS TC-PCT, the trade-off between the axial and lateral image resolution is the frame rate of the infrared camera. The images shown in the manuscript were taken at 104 Hz, which led to good spatial and axial resolution. However, higher infrared camera frame rates (from 104 Hz to 800 Hz) would have positively impacted the axial resolution especially harvesting more camera frames from the thermal transient decay signal, e.g. the first 200 ms. To further increase the penetration depth in MS TC-PCT imaging, one solution is applying an LFM chirp with lower starting frequency. For this, one would require a Q-switch trigger circuit and an automated customized LabVIEW program to adjust the timing of the Q-switch signal in relation to the flashlamp emission to produce LFM chirps. In most cases, this adjustment has to be done with respect to the Q-switch delay setting in microseconds. This value indicates the time between the command to fire the flashlamp and the command to fire the Pockels cell. Lastly, one limiting factor as the wavelength of the OPO increased was the reduction in pulse peak energy. For future research of a similar kind, optical tools can be introduced such as a converging lens to reduce the diameter of the laser pulse and enhance the pulse peak energy.
In summary, MS TC-PCT imaging of bacterial-induced dental caries and associated natural enamel defects in this work extended the single-wavelength TC-PCT hard-tissue imaging capabilities first introduced in [9] to a wide range of available wavelengths taking advantage of the concurrence of tissue-light scattering and absorption phenomena. For biological samples, the goal is to inspect “health defects” such as subsurface lesions located at various depths, and this requires the molecular specificity of light which necessitated the development of multispectral (MS) TC-PCT. It was shown that near-surface and deep subsurface features can be imaged, thereby making MS TC-PCT a very promising modality for further applications to hard tissue (dental and bone) imaging. In addition, the clinical MS TC-PCT imaging can be used for, but not limited to, absorption spectroscopic related detection of very early tumors, cancer detection, and treatment studies. Animal models can be used to elucidate cancer pathophysiology, identify novel therapeutic agents, and to study mechanisms of intrinsic and acquired resistance to targeted therapies.
Funding
CFI-JELF program (38794); The New Frontiers in Research Fund – Exploration (NFRFE-2019-00647); NSERC – Collaborative Research and Training Experience (496439-2017); Canada Foundation for Innovation (Research Chairs Program (950-230876)); Natural Sciences and Engineering Research Council of Canada (Discovery Grants Program (RGPIN-2020-04595)).
Acknowledgments
The authors gratefully acknowledge the Natural Sciences and Engineering Research Council (NSERC) Discovery Grants Program (RGPIN-2020-04595), the Canada Foundation for Innovation (CFI) Research Chairs Program (950-230876), the NSERC – Collaborative Research and Training Experience (CREATE) (496439-2017), the New Frontiers in Research Fund – Exploration (NFRFE-2019-00647) fund and the CFI-JELF program (38794) for financial support. The authors would also like to thank Dr. Pantea Tavakolian for her technical assistance in activation of the Surelite OPO.
Disclosures
The authors declare no conflicts of interest.
Data availability
Data underlying the results presented in this paper are not publicly available at this time but may be obtained from the authors upon reasonable request.
References
1. Oral Health, World Health Organization (2022).
2. E. A. Abou Neel, A. Aljabo, A. Strange, S. Ibrahim, M. Coathup, A. M. Young, L. Bozec, and V. Mudera, “Demineralization-remineralization dynamics in teeth and bone,” Int. J. Nanomed. 11, 4743–4763 (2016). [CrossRef]
3. J. R. Keenan and A. V. Keenan, “Accuracy of dental radiographs for caries detection,” Evid Based Dent 17(2), 43 (2016). [CrossRef]
4. E. B. Shokouhi, M. Razani, A. Gupta, and N. Tabatabaei, “Comparative study on the detection of early dental caries using thermo-photonic lock-in imaging and optical coherence tomography,” Biomed. Opt. Express 9(9), 3983–3997 (2018). [CrossRef]
5. Y.-S. Hsieh, Y.-C. Ho, S.-Y. Lee, C.-C. Chuang, J. Tsai, K.-F. Lin, and C.-W. Sun, “Dental optical coherence tomography,” Sensors 13(7), 8928–8949 (2013). [CrossRef]
6. H. E. Strassler and M. L. Pitel, “Using fiber-optic transillumination as a diagnostic aid in dental practice,” Compend. Contin. Educ. Dent. 35, 80–88 (2014).
7. J. Vaarkamp, J. J. Ten Bosch, E. H. Verdonschot, and S. Tranaeus, “Quantitative diagnosis of small approximal caries lesions utilizing wavelength-dependent fiber-optic transillumination,” J. Dent. Res. 76(4), 875–882 (1997). [CrossRef]
8. M. I. G. Ortiz, C. de Melo Alencar, B. L. F. De Paula, M. B. Magno, L. C. Maia, and C. M. Silva, “Accuracy of near-infrared light transillumination (NILT) compared to bitewing radiograph for detection of interproximal caries in the permanent dentition: A systematic review and meta-analysis,” J. Dent. 98, 103351 (2020). [CrossRef]
9. P. Tavakolian, K. Sivagurunathan, and A. Mandelis, “Enhanced truncated-correlation photothermal coherence tomography with application to deep subsurface defect imaging and 3-dimensional reconstructions,” J. Appl. Phys. 122(2), 023103 (2017). [CrossRef]
10. S. Kaiplavil and A. Mandelis, “Truncated-correlation photothermal coherence tomography for deep subsurface analysis,” Nat. Photonics 8(8), 635–642 (2014). [CrossRef]
11. P. Tavakolian, E. B. Shokouhi, S. Sfarra, G. Gargiulo, and A. Mandelis, “Non-destructive imaging of ancient marquetries using active thermography and photothermal coherence tomography,” Journal of Cultural Heritage 46, 159–164 (2020). [CrossRef]
12. S. Kaiplavil, A. Mandelis, X. Wang, and T. Feng, “Photothermal tomography for the functional and structural evaluation, and early mineral loss monitoring in bones,” Biomed. Opt. Express 5(8), 2488–2502 (2014). [CrossRef]
13. P. Tavakolian, S. Roointan, and A. Mandelis, “Non-invasive in-vivo 3-D imaging of small animals using spatially filtered enhanced truncated-correlation photothermal coherence tomography,” Sci. Rep. 10(1), 13743–10 (2020). [CrossRef]
14. ANSI Z136 American National Standard for Safe Use of Lasers (2007).
15. P. Tavakolian and A. Mandelis, “Perspective: principles and specifications of photothermal imaging methodologies and their applications to non-invasive biomedical and non-destructive materials imaging,” J. Appl. Phys. 124(16), 160903 (2018). [CrossRef]
16. R. Welch, K. Sivagurunathan, P. Tavakolian, K. Ngai, B. Huang, S. Abrams, Y. Finer, and A. Mandelis, “Detection of bacterial-induced dental caries using three-dimensional near-infrared photothermal imaging,” submitted 26(4), 046004 (2022).
17. . “ImageJ,” https://imagej.nih.gov/ij/.
18. R. S. Singh, M. O. Culjat, M. Lee, D. B. Bennett, S. Natarjan, B. P. Cox, E. R. Brown, W. S. Grundfest, and H. Lee, “Acoustical Imaging,” 31, 211–222 (2012).
