Sage Journals: Discover world-class research

Abstract

Detection of tooth caries using imaging at one or two NIR wavelengths is entering clinical practice, with imaging based on an increase in O-H overtone and combination features associated with increased porosity and water content in decayed or demineralised teeth, and/or change in light scattering with change in tooth porosity. Further consideration of the spectroscopy behind this application is warranted, given details of a NIR spectroscopy-based model for tooth water content has not been previously reported. A partial least squares regression model for tooth water content was developed using a second derivative pretreatment and five factors, achieving a cross validation R² of 0.91 and RMSECV of 0.4 % w/w, with model weightings consistent with reliance on O-H features. This result indicates that caries detection is likely based on (water) absorption rather than scattering features.

Keywords

hydroxyapatite vibrational spectroscopy chemometrics

Introduction

Although apparently solid, calcified tissues such as teeth are porous and can absorb water (to 12% by weight).¹ Porosity is increased in caries affected or acid exposed teeth,² with these effects ameliorated in fluoride treated teeth.³ Infrared and near infrared (NIR) spectra of intact teeth therefore carriy information on pore water content through O-H absorption and light scattering.

These phenomena are the basis of several dental imaging applications for assessing tooth porosity and thus identifying lesions on a tooth, using illumination at one or two NIR spectral region wavelengths.⁴ These imaging applications initially focussed on use of illumination around 1300 nm (7634 cm⁻¹) based on the observation that the scattering coefficient at this wavelength was increased by more than two orders of magnitude by natural and artificial enamel demineralisation.⁵ Imaging at this wavelength has been used for qualitative assessment of the severity of occlusal carious.⁶

Consideration was later given to use of water absorption features, for example, at around 5128 cm⁻¹(1950 nm)⁷, 6849 cm⁻¹ (1460 nm),⁸ 10,427 cm⁻¹ (959 nm)^9,10 or 11,765 cm⁻¹ (850 nm),¹¹ associated with the first, second and third overtone regions, respectively, of the O-H stretching vibration. Active lesions dry quicker than arrested lesions or healthy areas of a tooth because of differences in pore size, with change at 1950 nm (5128 cm⁻¹) shown to be more pronounced than at other wavelengths in the 1500–2060 nm (4854–6667 cm^-1) range.⁷ Imaging using two wavelengths, for example, use of absorption at 1440 nm (6944 cm⁻¹) normalised to that at 1090 nm (9174 cm⁻¹), has also been used to map the spatial distribution of lesions in teeth.¹² More recently, imaging at 1950 (5128 cm⁻¹) and 1300 nm (7634 cm⁻¹) during forced air drying of the tooth over approximately 60 s was used to generate dehydration curves and recommended for improved discrimination between active and arrested lesions.¹³

Commercial providers have begun supplying imaging equipment for clinical practice use, with operation in the Herschel region (<1100 nm) being used because of the lower cost of such instrumentation. For example, the VistaCam iX HD (Dürr Dental SE, Bietigheim-Bissingen, Germany) operates using 850 nm (11,765 cm⁻¹) illumination.

However, while the imaging applications using water absorption features are used in qualitative evaluation of tooth lesions, a quantitative relationship between absorbance and tooth water content has not been described. A partial exception is a study in which NIR hyperspectral imaging (950–1600 nm/6250–10,526 cm⁻¹) was used in estimation of the rate of water evaporation from teeth¹⁰ (either untreated or subject to a demineralising treatment of pH 4.3 lactic acid). A PLS model was used in estimation of water content; however, no details of the model were provided.

This Technical Note addresses this gap, providing detail of a PLS model for tooth water content and providing the dataset for public use. This is intended to underpin existing work on use of water features in assessment of tooth characteristics.

Materials and methods

Sample material

Sound teeth (n = 23, central and lateral incisors, canines, premolars and molars) from 13 donors were collected from private- and public dental clinics across Central Queensland, Australia under ethics approval (Central Queensland University #21089 and Queensland Health #48361). The teeth, which were extracted for periodontal and orthodontic reasons, were devoid of decay, visible cracks and restorative materials. The teeth were recorded as vital or non-vital at the time of extraction by the treating dentist. Teeth were cleaned using a soft toothbrush and distilled water, air-dried for 7 days and then stored in plastic specimen jars.

Spectra acquisition and treatments

Near infrared spectra (4003–9886 cm⁻¹; resolution 16 cm⁻¹; 32 scans per spectrum) were acquired of intact human teeth using the integrating sphere chamber (10 mm diameter) of a Nicolet Antaris FT-NIR analyser (Thermo Fisher Scientific, Waltham, MA, USA), employing the in-built reference, with samples in an air atmosphere and at room temperature, following the method of Pretorius et al.¹⁴

Air-dried teeth were hydrated by immersion in distilled water for 1, 2, 5, 8, 10, 15, 30, 45, 60, 75, and 90 min. After each interval, surface moisture was removed with absorbent paper prior to spectral acquisition and gravimetric measurement. Dehydration was then induced by storing teeth in vials with dry silica gel at room temperature, with measurements repeated after 15 min and 24 h. Final measurements were obtained following oven-drying at 65°C for 6 weeks.

Data analysis

Absorbance spectra were exported to The Unscrambler® (X 10.5.1, Camo, Norway) software for principal component analysis (PCA) and partial least squares (PLS) regression analysis using the NIPALS algoritm. PLS calibration development was based on a 20-group cross validation, approximating to a leave-one-out procedure. The selection of the optimal number of factors (5) was based on minimisation of RMSECV value.

Results and discussion

Tooth water content

Upon immersion in water, air-dried teeth absorbed approximately 3.7% of their initial weight within 2 minutes, plateauing at 11.5% w/w after 24 h, while a 24 h desiccator treatment removed approximately one half of the absorbed water and oven drying (65°C) resulted in a further decrease in weight to less than the initial air-dry weight. Thus, as expected from the reported porosity of teeth,¹⁵ teeth contain spectroscopically significant levels of moisture.

Spectra

The absorption of water by a tooth was associated with a general absorbance increase across the spectrum consistent with a decrease in light scattering by the tooth and thus decrease in diffuse reflectance, and change in water-related absorption features, for example, from 6650-7270 cm⁻¹ and 4740–5400 cm⁻¹ (Figure 1). The small sharp feature at 6983 cm⁻¹ associated with O-H of structural water¹⁶ was clear in oven dried samples but diminished in hydrated samples that contained larger amounts of loosely bound water.

Figure 1.

(a) Absorbance and (b) Second derivative of absorbance spectra of a sound upper third molar tooth before (Tooth Pre) and after immersion in water for 90 min (H₂O), and after oven drying at 65°C for 6 weeks. (c) Difference absorbance spectra for teeth after immersion in water for 24 h and oven drying at 65°C for 2 and 6 weeks, relative to pre immersion.

Air-dried teeth that were vital at the time of extraction were distinguished from those that were non-vital in a PCA score plot based on second derivative of absorbance spectra (Figure 2), although water content was similar for the two groups (mean ± SE of 1.0238 ± 0.0129% w/w for vital teeth and 1.0195 ± 0.0136% w/w for non-vital teeth). The loadings of the PCA model contained features attributable to water features, at around 4800–5200 cm⁻¹ and 6800–7200 cm⁻¹. This result is consistent with the higher porosity of non-vital teeth.¹

Figure 2.

PCA scores plot based on second derivative (Savitzky-Golay 2^nd polynomial 31 smoothing point) and standard normal variate (SNV) treated absorbance data of air-dried teeth that were either vital or non-vital at the time of extraction. The percentage of explained variation for each PC is shown in brackets. Red circles represent teeth that were non-vital at the time of extraction, cyan circles represent teeth that were vital at the time of extraction.

Water model

Inclusion of oven dried samples in the model training set led to poor model performance, with a five factor PLS regression model achieving a R²_cv of 0.72 and RMSECV of 1.0% w/w (Figure 3(a)). This result is consistent with the involvement of a different state of water, viz., structural water as opposed to loosely bound water, when adding oven dried samples.

Figure 3.

(a) Scatter plot of predicted vs reference tooth weight as a % of air dry weight using a (5 factor) PLS regression model and second derivative of absorbance data for air dried and heat dried teeth (n = 23). Calibration results are shown in blue, cross validation results in red. (b) As for (a), but excluding heat dried samples.

Following assessment of a range of pretreatments, a pretreatment of second derivative Savitzky-Golay (2^nd polynomial order, 31 points left and right) was used. The explained variance for X across the first five PLS factors was: 36, 38, 17, 7 and 1%, and for Y: 57, 9, 13, 9 and 9%, with a cumulative explained variance of 99% for X and 97% for Y. The five factor PLS regression model achieved a R²_cv of 0.91 and RMSECV of 0.4% w/w (Figure 3(b)). By way of comparison, a PLS regression model based on absorbance data achieved R²_cv of 0.53 and RMSECV of 1.0% w/w (data not shown), while R for a univariate correlation using absorbance at 1950 nm (5130 cm⁻¹) and 1300 nm (7691 cm⁻¹) was 0.57 and 0.20, for this dataset. The predictive performance on the subsets of vital and non-vital teeth were similar, with non-vital teeth yielding a RMSECV of 0.184% w/w and R² of 0.979, while vital teeth produced a RMSECV of 0.251% w/w and R² of 0.960.

The model beta coefficients and loadings for factor 1 (Figure 4) demonstrated weighting around 5300, 5400 and 7200 cm⁻¹, interpretable as associated with absorption features of water¹¹ and consistent with reliance on water absorption features rather than light scattering per se.

Figure 4.

(a) Beta coefficients and (b) loadings plot for Factor 1 of a five factor PLS regression model on tooth weight (% of air-dry weight), based on second derivative of absorbance data. Arrows indicate the position of absorption bands associated with water (at approximately 5300, 5400, and 7200 cm⁻¹).

Conclusion

Details of a PLS model for tooth water content have been provided, representing, to the best of knowledge, the first report of model performance statistics for this application. This result provides a foundation for existing work on the use of water features in assessment of tooth characteristics and is consistent with a greater contribution from water absorbance features rather than scattering features in imaging techniques for caries detection.

Several limitations to the current study exist: (i) the relatively small sample size (n = 23), which although adequate for demonstrating proof-of-concept, constrains broader generalisation of the model; (ii) only sound extracted teeth were analysed, with exclusion of carious or restored teeth, which may exhibit different spectral behaviours; (iii) hydration and dehydration were induced under laboratory conditions, which may not fully replicate the dynamics of water exchange in vivo.

Future work should include validation of the model on larger and more diverse datasets, including carious and restored teeth. Extension of this work to in situ or in vivo measurements will be required for evaluating clinical utility. Finally, comparison of spectral modelling approaches across different NIR regions and integration with imaging modalities may strengthen predictive performance.

Supplemental Material

Supplemental Material - Estimation of the water content of human teeth using near infrared spectroscopy

Supplemental Material for Estimation of the water content of human teeth using near infrared spectroscopy by N. E. Pretorius, A. Forrest, and K. B. Walsh in Journal of Near Infrared Spectroscopy

Footnotes

Acknowledgements

The authors thank Prof Mark Tennant for academic advice, and Dr Surya Bhattarai for assistance with the use of Sigmaplot® software.

ORCID iDs

Nellie Elizabeth Pretorius

Kerry Brian Walsh

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The authors gratefully acknowledge financial support provided by the Australian Government through the Commonwealth Research Training Program.

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Supplemental Material

Supplemental material for this article is available online.

References

Papa

Cain

Messer

. Moisture content of vital vs endodontically treated teeth. Dent Traumatol 1994; 10: 91–93.

Sodata

Peerapattana

. Application of near infrared spectroscopy with chemometrics for qualitative and quantitative dental caries assessment. Vib Spectrosc 2020; 111: 103170.

ADA . Policy statement 2.2.1 - fluoride use. Council

(ed). Australian Dental Association, 2023.

Chang

N-YN

Jew

Fried

. Lesion dehydration rate changes with the surface layer thickness during enamel remineralization. Proc SPIE Int Soc Opt Eng 2018; 10473: 104730D.

Darling

Huynh

Fried

. Light scattering properties of natural and artificially demineralized dental enamel at 1310 nm. J Biomed Opt 2006; 11: 034023.

Lee

Darling

, et al. Nondestructive assessment of the severity of occlusal caries lesions with near-infrared imaging at 1310 nm. J Biomed Opt 2010; 15: 047011.

Tressel

Abdelaziz

Fried

. Dynamic SWIR imaging near the 1950-nm water absorption band for caries lesion diagnosis. J Biomed Opt 2021; 26: 056006.

Sahyoun

. Short-wave infrared optical imaging techniques for the assessment and quantification of dental disease. Rutgers The State University of New Jersey, School of Graduate Studies, 2020.

Tetschke

Kirsten

Golde

, et al. Application of optical and spectroscopic technologies for the characterization of carious lesions in vitro. Biomed Tech 2018; 63: 595–602.

10.

Usenik

Bürmen

Fidler

, et al. Near-infrared hyperspectral imaging of water evaporation dynamics for early detection of incipient caries. J Dent 2014; 42: 1242–1247.

11.

Kolmas

Marek

Kolodziejski

. Near-infrared (NIR) spectroscopy of synthetic hydroxyapatites and human dental tissues. Appl Spectrosc 2015; 69: 902–912.

12.

Zakian

Pretty

Ellwood

. Near-infared hyperspectral imaging of teeth for dental caries detection. J Biomed Opt 2009; 14: 064047.

13.

Wycoff

Zhu

Fried

. In vitro assessment of lesion activity using simultaneous time-resolved reflectance imaging at 1300 and 1950 nm. Laser Med Sci 2024; 39: 223.

14.

Pretorius

Subedi

Forrest

, et al. Near infrared spectroscopy in the characterisation of intact human teeth inside and outside custody bags. J Near Infrared Spectrosc 2021; 29: 102.

15.

Mann

Dickinson

. Nanomechanics, chemistry and structure at the enamel surface. Monogr Oral Sci 2006; 19: 105–131.

16.

Berzina-Cimdina

Borodajenko

. Research of calcium phosphates using Fourier transform infrared spectroscopy. In: Infrared spectroscopy-materials science, engineering and technology. Books on Demand, 2012; Vol. 12, pp. 251–263.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.70 MB