Validity and Reliability of Intraoral Optical Coherence Tomography and Bitewing Radiography for Detecting Approximal Carious Lesions

Caries is the most prevalent chronic human disease worldwide [1, 2]. There is an increase in carious lesions from adolescence onward, especially on the approximal surfaces [3‒10].

Assessment of approximal caries lesions remains a diagnostic challenge. The established tools, visual-tactile examination, and radiography show clear limitations, especially in detecting early approximal carious lesions [11‒14]. Detection of demineralized and porous enamel and monitoring lesions with high spatial resolution are requirements for effective monitoring [15, 16]. Visual inspection and bitewing radiography for detecting approximal carious lesions depends significant on the diagnostic threshold, which vary across different studies and clinical practices [17‒19]. In everyday practice, it is impossible to reliably assess the long-term activity of caries lesions with a single-visit visual and tactile examination [20]. Meta-analyses demonstrated a moderate to high specificity (SP) but low sensitivity (SE) for visual (SP: 0.9–0.99, SE: 0.25–0.35) and radiological (SP: 0.70–0.97, SE: 0.24–0.61) examination, especially in the detection of initial approximal caries [17, 21‒24]. The low SE leads to the progression of demineralization up to the middle third of the dentin when the lesion is already clearly visible in radiographs [25]. Therefore, dentists mostly use a combination of bitewing radiographs and visual-tactile examination, increasing the SE to 79% [26‒28]. Nevertheless, noninvasive diagnostic methods with higher reliability are desirable, especially considering early approximal caries lesions.

Optical coherence tomography (OCT) is a noninvasive imaging technique that generates 3D images of biological structures [29], and is based on white light interferometry. While swept-source OCT (SS-OCT) uses a laser as a light source that rapidly sweeps the wavelength over a range of wavelengths in combination with a photo detector, SD-OCT uses a broadband light source and a spectrometer to capture the interference patterns of light reflected from different tissue layers and the reference arm [30]. With SD-OCT at the central wavelength of 1,310 nm ± 120 nm, object surfaces and structures contained therein with different refractive indices can be detected up to a penetration depth of approx. 2.5 mm (tooth enamel) with high lateral (≤15.0 μm) and axial (6.6 μm, water) resolution in the micrometer range. The high imaging speed of the OCT systems used enables real-time imaging during the detection of caries. Numerous studies have shown the potential of OCT in early caries diagnosis [30‒35]. Handheld OCT probes, which allow the examination of occlusal and approximal tooth surfaces up to the third molars, have recently been developed and tested in vitro [36‒40] and in vivo [39‒41]. Preliminary caries diagnosis studies revealed higher SE (81–95%) for OCT compared to visual inspection (74–88%) [36, 37, 42] and higher SE (89%) as well as SP (73%) compared to radiography (SE 73%, SP 63%) [41, 43]. An in vivo assessment of approximal carious lesions with a handheld intraoral OCT probe compared to visual-tactile assessment, radiography, and fiber-optic transillumination has been described [39]. However, the validity and reliability of approximal caries detection have not yet been investigated.

The aim of this in vitro study was to evaluate the reliability and validity of the intraoral OCT probe in the assessment of initial lesions at approximal surfaces and compare it to digital radiography (DR). Since OCT has been described to show higher SE and SP in detecting initial caries lesions, it was expected that this could be demonstrated in the present study setting.

Forty extracted human premolars and molars with unrestored healthy or carious approximal surfaces were embedded in C-silicone (Henry Schein Inc., Melville, NY, USA) according to their natural arrangement and mounted in a patient-equivalent simulator (PK-6, Frasaco GmbH, Tettnang, Germany) (Fig. 1a). Fifty-four unrestored approximal surfaces with ICDAS ll scores 0–3 (Table 1) were orthogonally photographed (Nikon D7000, Nikon AF-S NIKKOR 105 mm, Nikon GmbH, Düsseldorf, Germany) and imaged using DR (see 2.1) and a prototype of an intraoral OCT probe (SD-OCT, see 2.2). The caries status of the approximal tooth surfaces was verified by X-ray microtomography (µCT; 85 kV, 118 μA, filter Al/Cu, rotation step 0.2°, averaging 4, pixel size 2.6 μm, random movement 10; Skyscan 1172–100-50, Bruker MicroCT, Kontich, Belgium) and light microscopy (Stemi 2000 C, Carl Zeiss Microscopy GmbH, Jena, Germany; sectioning of teeth up to the maximum extent of the lesion, Leitz 1,600 sawing-microtome, Ernst Leitz Wetzlar GmbH, Wetzlar, Germany). The extension of the approximal lesions in µCT and light microscopic images were scored according to the scoring system for OCT images, a modification of the scoring system for radiographs (Table 2). The DR and OCT cross-sectional images (B-scans) were analyzed by five dentists with limited to extensive experience in evaluating radiographs and OCT B-scans.

At a dental unit (TENEO, Sirona Dental Systems GmbH, Bensheim, Germany), digital bitewing radiographs were taken with a Heliodent DS Dental X-ray unit using the paralleling technique (Fig. 1b). The technical settings were 7 mA and 60 kV (Dentsply Sirona Deutschland GmbH, Bensheim, Germany; bitewing film holder, Kentzler-Kaschner Dental GmbH, Ellwangen/Jagst, Germany) using an image plate, and sensor size of 3 × 4 cm, Dürr Dental AG, Bietigheim-Bissingen, Germany). The images were scanned with the VistaScan Mini Plus (Dürr Dental AG, Bietigheim-Bissingen, Germany) and analyzed using the DBSWIN software (version 5.15.1, Dürr Dental AG, Bietigheim-Bissingen, Germany). Approximal areas from the first premolar’s distal surface to the second molar’s mesial surface were evaluated using the modified scoring system described by Pitts [44, 45] (Table 2).

The spectral-domain OCT (SD-OCT) system Telesto-ll SP 21 with a modified handheld scanner (OCTH-1300NR-SP6, Thorlabs GmbH, Dachau, Germany; endoscopic probe, Medizinisches Laserzentrum Lübeck GmbH, Lübeck, Germany) permits chairside intraoral scanning of healthy and carious tooth surfaces [39]. Approximal tooth surfaces were imaged with the handheld probe (Fig. 1d) from the occlusal with a maximum field of view 8 mm × 8 mm × 3.5 mm (central wavelength ± bandwidth: 1,550 nm ± 115 nm, maximum pixel size 800 × 400 × 1,024, lateral resolution 30 μm, A-scan rate 28 kHz). OCT B-scans were obtained in real-time. The B-scan images were scored in adaptation to the modified radiography scoring system (Table 2).

Five clinicians (rater 1, 2, 3, 4, 5) with high level (rater 1, 9 years) to low level (rater 5, 0 years) of expertise in X-ray and OCT B-scan evaluation were calibrated by an evaluator (C.R.) with a high level of expertise in the assessment of OCT B-scans. Calibration for evaluating the OCT sectional images was standardized for all raters and took half an hour. In this time, an OCT expert with high level of expertise in assessment of OCT images first explained the different lesion signals in OCT images. The expert explained the different signals of lesions (enamel/dentin), artifacts, and structures of the tooth. Various OCT cross-sectional images with all these structures were then evaluated, the evaluations compared and discussed together. The decision-making process of the rater was repeated two times.

The clinicians analyzed the X-ray and OCT images independently after blinding the photographs, the X-rays, and the OCT scans were analyzed. They repeated the assessments after two to 4 weeks. The X-ray and OCT images were scored in the same room with a high-resolution monitor (Color Edge CG 248, EIZO, Hakusan, Japan) certified for X-ray diagnosis. Intra- and interrater agreement was calculated with weighted Cohen’s and Fleiss’ kappa to assess rater reliabilities (IBM SPSS Statistics 27 for Windows; IBM Corp., Armonk, NY, USA). The assessment systems were harmonized to evaluate the agreement between X-ray and OCT assessment (Table 3). To determine the diagnostic accuracy of OCT and DR, SE, and SP were calculated at the two time points for the five raters. The results were validated with X-ray µCT: a technical assistant with high experience in evaluation of X-ray µCT images scored expansion of carious lesions analog to the OCT and DR image scoring systems (Table 3).

The intrarater agreements (kappa values) for the five examiners in DR and OCT assessment showed a slightly higher agreement for OCT than for DR in 4 out of 5 raters (Table 4). For DR, the rating agreed fair to substantial (0.36–0.78; p < 0.05), and for OCT, the rating agreed moderate to substantial (κ: 0.53–0.77; p < 0.001). The comparison of both methods showed slight to moderate agreement at both points in time (κ: 0.16–0.55; p < 0.05).

Interrater agreement for DR and OCT are given at two different times (Table 5). The Cohen’s kappa values were higher for OCT than for DR in 9 out of 10 interrater comparisons. The strength of pairwise agreement at both assessment times ranges from slight to substantial for DR (κ: 0.08–0.74, Fleiss’ κ: 0.23/0.24; p < 0.001). For OCT, the pairwise agreement was fair to substantial (κ: 0.27–0.62; Fleiss’ κ: 0.23/0.18; p < 0.001).

SE and SP for detecting approximal carious lesions are presented in Table 6. For all raters, OCT showed higher values of SE (0.27–0.91) and similar values of SP (0.25–1) than DR (0.16–0.82/0.33–1). SP values were higher for OCT than DR in the experienced raters 1 and 2. OCT resulted in higher diagnostic accuracy than DR (Fig. 2-7).

The approximal carious lesions could be visualized using X-ray µCT. The agreement between µCT and histology was almost perfect (κ: 0.82; p < 0.001), and between DR or OCT and µCT, it was slight to moderate (Table 7). For each rater, the agreement between DR and µCT (κ: 0.05–0.45; p < 0.05) was lower than between OCT and µCT (κ: 0.18–0.50; p < 0.05).

This study compares the validity and reliability of a noninvasive intraoral OCT probe and digital bitewing radiography in detecting and assessing early approximal carious lesions. The results demonstrate, for the first time in a clinical simulation setting, that OCT is more sensitive than DR in detecting early approximal carious lesions without losing SP or reproducibility.

The SE of radiographic caries detection is low for all lesions, particularly those in their initial stages. Therefore, the values for approximal caries (0.36 [0.24/0.49]) are even lower than for occlusal carious lesions (0.56 [0.53/0.59]) [22]. It is, therefore, understandable that there is a desire for a high-SE caries diagnostic method. However, this should not be at the expense of diagnostic quality in healthy tooth surfaces (SP). To address this issue, we investigated healthy approximal surfaces and those with early carious lesions up to ICDAS ll score 3. The detection and monitoring of these early lesions is essential for early non- or minimally invasive intervention to prevent caries progression toward more advanced defects.

OCT offers a considerable gain in information in the detection of enamel lesions, as the near-infrared light is strongly scattered in the lesion body compared to healthy enamel and is attenuated and scattered more strongly in the dentin than in the enamel [46]. In the detection of caries under composite restorations, the reliability of SS-OCT was higher than that of DR [47]. The present study confirmed this result for SD-OCT in early approximal caries detection. In four out of five examiners, intrarater reliability was higher with OCT than with DR. This is remarkable as visual inspection (ICDAS II), and bitewing radiography are currently considered the most reliable methods for detecting approximal caries [48].

The validity of SD-OCT and DR were calculated based on a microtomographic validation standard. In this study, histology and X-ray µCT were compared in advance to determine their agreement. We had deemed this necessary to rule out the possibility that early enamel lesions might not be adequately represented in the µCT B-scan. A meta-analysis revealed a low risk of bias when using X-ray µCT, especially when identifying true negatives worked well to excellent [49]. Our findings confirmed this with the almost perfect agreement between µCT and histology (κ: 0.82; p < 0.001). There was greater agreement of lesion extent with OCT (κ: 0.18–0.50) than with radiography (κ: 0.05–0.45). However, when sectioning the carious lesions at the largest extension, the plane does not always match precisely the plane imaged with OCT due to its three-dimensional morphology. Further, the depiction of the approximal lesions in DR images was disturbed by superimpositions and summation effects, limiting image quality compared with µCT images. The radiographs were scored using the modified scoring system described by Pitts [44, 45]. We achieved a higher spatial resolution with OCT than DR, prompting us to use more refined scaling for analyzing OCT images. In perspective, the more precise assessment of lesions also enables more accurate monitoring of initial caries lesions, which is a clear advantage of OCT compared to other available diagnostic methods.

The results for SE and SP generally agree with those of previous studies that used OCT intraoral probes to assess sound and carious approximal surfaces [38, 41, 43]. Using SS-OCT, other authors showed a higher SE compared to radiography for the detection of approximal lesions in the clinical setting [41]. In this study, in addition to the higher SE of the intraoral SD-OCT probe compared to DR, it was found that OCT also achieves higher SP values than DR, especially in the more experienced of the five examiners. Similarly, Nakagawa et al. [36] showed a high SE in detecting initial approximal caries lesions with OCT for investigators with three and 9 years of OCT experience but a lower SP for examiners with less experience. The healthy approximal surfaces on the OCT images were more often incorrectly classified as carious by the less experienced evaluators. Not surprisingly, from a training perspective, with increasing experience in OCT, it seems easier to distinguish carious lesions from enamel cracks or other structural changes.

In recent years, more and more methods for the early detection of caries lesions have been investigated, e.g., laser-fluorescence (DIAGNOdent™ pen/cam [50‒53]) and light-based methods (Qraycam [54]). Compared to the DIAGNOdent™ pen (SE 0.6–0.88, SP 0.2–0.98 [55]) and the DIAGNOcam™ (SE 0.8–0.92, SP 0.38–0.81 [50, 51, 56]), similar values resulted in this study with the intraoral OCT probe used (SE 0.27–0.91, SP 0.25–1). When comparing different diagnostic methods, attention should be paid to the physically defined differences in the visualization of lesions and defects, as this cannot be directly derived from the validity values. OCT may have an advantage in the particularly precise and reproducible visualization of early carious lesions and provides three-dimensional image stacks, allowing detailed quantification. Future studies should take a differentiated look at the advantages and limitations of the new diagnostic methods for varying diagnostic scenarios. At present, no single method is best suited for a wide range of indications in caries diagnosis and, specifically, in approximal caries diagnosis. Targeted use depends on the diagnostic focus, and the combination of methods with different target parameters seems more promising for improving the quality of caries diagnosis.

The bright signals in SD-OCT images corresponded more closely to the carious lesion extent in µCT images than to the approximal brightening in digital radiographs. Compared to DR, OCT showed greater accuracy in assessing approximal tooth surfaces, especially if the teeth are in a closed row and the approximal space is narrow. The intraoral OCT data suggest that OCT is a valid and reliable tool for the early detection and assessment of non-cavitated approximal carious lesions. OCT could contribute to more accurate and effective caries monitoring, particularly when nonionizing radiation imaging is desirable or required. Calibration of the users is essential, as is sufficient experience in interpreting OCT images. The findings obtained in the clinic-oriented simulation now need to be confirmed in in vivo studies.

We sincerely acknowledge the excellent technical support by three clinicians (Olivia Giesa, Julia Möller, Johanna Becker; University of Leipzig, Department of Cariology, Endodontology and Periodontology).

This study was conducted in line with the Declaration of Helsinki and approved by the Ethics Committee of the Medical Faculty at the University of Leipzig, Approval No. 299-10-04102010. This study was an in vitro study without human participants. Written informed consent was obtained from participants prior to the study for the anonymized collection of extracted teeth in dental practices (reference number 299-10-04102010).

The authors have no conflicts of interest to declare.

This research was funded by the German Federal Ministry of Economic Affairs and Energy (ZIM/AIF-ZF4148701CR5), the European Regional Development Fund/Saxon State Ministry of Science and the Arts (EFRE/SMWK Grant 100175024), the German Research Foundation/Saxon State Ministry of Science and the Arts (DFG/SMWK 376/7-1 FUGG), and the University of Leipzig within the Open Access Publishing program.

MS: writing – original draft, visualization, data curation, methodology, investigation, and formal analysis. HS: writing – review and editing, conceptualization, supervision, funding acquisition, and formal analysis. CR: data curation, supervision, and methodology. JS: writing – review and editing, investigation, and supervision. ESK: writing – review and editing and supervision. RH: writing – review and editing, conceptualization, supervision, and funding acquisition.

All data generated or analyzed during this study are included in this article. Further inquiries can be directed to the corresponding author.