Near-infrared imaging in orthodontic intraoral scanners for early interproximal caries detection

Introduction

Intraoral scanners commonly used in orthodontic offices now offer near-infrared imaging (NIRI) technology, advertised as a screening tool to identify interproximal caries. This study aimed to evaluate the reliability and validity of NIRI detection of interproximal carious lesions in a common intraoral scanner (iTero Element 5D; Align Technology, San Jose, Calif) with and without bitewing radiograph complement, compared with a microcomputed tomography (micro-CT) reference standard.

Methods

Extracted human posterior teeth (premolars and molars) were selected for early (noncavitated) interproximal carious lesions (n = 39) and sound control surfaces (n = 47). The teeth were scanned via micro-CT for evaluation by 2 blinded evaluators using consensus scoring. The teeth were mounted to simulate anatomic interproximal contacts and underwent a NIRI scan using iTero Element 5D and bitewing radiographs. Two trained, calibrated examiners independently evaluated (1) near-infrared images alone with clinical photograph, (2) bitewing radiograph alone with clinical photograph, and (3) near-infrared images with bitewing radiograph and clinical photograph in combination, after at least a 10-day washout period between each evaluation.

Results

Interrater reliability was highest for NIRI alone (k = 0.533) compared with bitewing radiograph alone (k = 0.176) or in combination (k = 0.256). NIRI alone showed high specificity (0.83-0.96) and moderate sensitivity (0.42-0.63) compared with a micro-CT reference standard. Dentin lesions were significantly more reliably detected than enamel lesions.

Conclusions

After rigorous training and calibration, NIRI can be used with moderate reliability, high specificity, and moderate sensitivity to detect noncavitated interproximal carious lesions.

Highlights

•

Evaluate near-infrared imaging interproximal caries detection in iTero 5D scanner (Align Technology, San Jose, Calif).
•

In vitro models allow for anatomic simulation and microcomputed tomography reference standards.
•

We compared near-infrared imaging and bitewing radiographs alone and in combination.
•

Near-infrared imaging had the highest reliability, high specificity, and moderate sensitivity.

Patients undergoing orthodontic treatment with fixed appliances are at an elevated risk of caries, including, but not limited to, white spot lesion development. A recent orthodontic consensus paper stated that “an initial comprehensive assessment should be performed to evaluate …oral health” and that “caries risk assessment should be performed before and routinely during orthodontic treatment.” A thorough caries examination is time intensive and is complemented by bitewing radiographs (BWR), especially for interproximal surfaces, which are not routinely captured in the orthodontic clinic. However, some limitations of BWRs include their low sensitivity for early enamel carious lesions and their limited interval frequency because of risks associated with ionizing radiation. ^, A novel method for interproximal caries evaluation that is readily available in the orthodontic clinic is near-infrared imaging (NIRI) technology, which is now found in intraoral scanners. Potential benefits of NIRI intraoral scans include its nonionizing radiation, comfortable patient experience, and frequency of capture as part of orthodontic records.

Intraoral scans have become increasingly common in orthodontics as clear aligner therapy increases, patient preferences shift, and orthodontists realize the benefits of efficient storage and retrieval, ease of transferability, durability, and decreased laboratory time. Beginning with the iTero Element 5D scanner (Align Technology, San Jose, Calif), several digital intraoral scanners include NIRI technology, advertised as an enhanced aid for detecting enamel defects, including interproximal caries, fluorosis, craze lines, and to distinguish staining from noncavitated early carious lesions. NIRI technology sends near-infrared wavelength light (850 nm) onto the tooth and uses a charge-coupled device sensor to capture images to be interpreted by the clinician. Given the use of intraoral scanning as part of orthodontists’ standard initial records, orthodontists may have the potential to use NIRI to screen for interproximal caries or caries risk before initiating orthodontic therapy.

NIRI is purported to identify caries-active patients, high caries-risk patients, or lesion development after orthodontic treatment has started. Whether or not orthodontists should use the information provided by NIRI depends on the reliability and validity of NIRI in detecting interproximal caries. Although significant research has been completed to evaluate NIRI validity in other devices, there is less evidence in the literature regarding the performance of NIRI in intraoral scanners. One in vitro study using NIRI technology (DIAGNOcam) in the TRIOS 4 intraoral scanner (3Shape A/S, Copenhagen, Denmark) found overall good diagnostic performance, although reduced sensitivity for initial carious lesions compared with a histologic gold standard. Another in vitro study evaluated iTero Element 5D scanners, more commonly found in orthodontic offices, and found low sensitivity and high specificity for NIRI evaluation of enamel and dentin interproximal caries against microcomputed tomography (micro-CT) reference standard. This study limited its evaluation to 2 adjacent teeth with a single interproximal contact and captured BWRs free of interproximal contacts, perhaps increasing the risk of bias. A recent in vivo study found that NIRI and BWR used in combination significantly increased interrater reliability, although a reference standard was not feasible in this study. Another in vivo study found that examiner training significantly affects outcomes using NIRI from intraoral scanners. However, further work is required to validate NIRI in vitro to both allow for a micro-CT reference standard and simulate anatomic interproximal contacts for evaluation by BWR and NIRI and, in combination, for interproximal caries detection. This would allow for a standardized assessment of sensitivity and specificity, in addition to interrater reliability assessment, as previously discussed.

Identifying a caries-suspicious surface on an NIRI scan is typically followed up with a BWR. An interesting clinical question is whether using NIRI and BWR in combination will increase sensitivity and specificity compared with either technology alone. This study aimed to evaluate the reliability and validity of NIRI interproximal caries detection in posterior teeth using iTero Element 5D as a standalone method and, in conjunction with BWRs, using micro-CT as a reference standard.

Material and methods

An overview of the methods is shown in Figure 1 . Extracted human posterior teeth were selected by visual-tactile examination using International Caries Classification and Management System criteria for sound control surfaces (n = 47) and noncavitated interproximal smooth surface lesions (n = 38). The power of the sample size was confirmed using power calculations with α = 0.05, and all values were >0.85, which identified the need for at least 16 teeth in each category. The selected teeth were excluded if they had existing restorations, fractures, or developmental defects, including fluorosis. Similar to a protocol described by our laboratory, the teeth were initially stored in 0.2% thymol solution at 4°C, cleaned via hand scaling, polished via Robinson brush (TPC Advanced Technology, City of Industry, CA) and deionized water (DIW), rinsed, and finally stored in DIW. The storage medium was changed every 3 days for 3 cycles while still being stored at 4°C. This study was approved by the Institutional University of North Carolina – Chapel Hill Office of Human Research Ethics (study no. 18-1654) and was classified as nonhuman subjects research, designated as exempt given that it does not constitute human subjects research.

The surfaces (n = 85) were scanned with a high-energy micro-CT (Scanco μCT 40, Scanco Medical AG, Bruttisellen, Switzerland) using a similar protocol to that used by our laboratory previously. The acquisition parameters are defined as follows: 70 kV, 114 μA, isotropic voxel size of 16 μm, scanned in air, with 250 projections per slice, and an integration time of 200 milliseconds. After the acquisition, the data are reconstructed with Scanco software (Scanco Medical AG) using a filtered back projection algorithm.

Two evaluators (A.M.H. and P.P.) independently examined all surfaces using MicroView Standard (version 2.5.0, Parallax Innovations; GE Medical Systems, Milwaukee, Wis). Each tooth was imported as an individual file, and a 3-dimensional reconstruction was generated using the Isosurface feature. The reconstruction was used to standardize the orientation for evaluation. The 2-dimensional slices were moved through the 3-dimensional reconstruction to evaluate the surface of interest (left or right interproximal). Surfaces of interest were evaluated in the sagittal and axial planes of space, scrolling the 2-dimensional slice searching for demineralization. If demineralization was identified, the stack was scrolled through to identify the deepest portion of the lesion for classification.

The 2 evaluators examined each surface independently using an adaptation of standard criteria: absence of demineralization (0), demineralization limited to the outer half of enamel (1), demineralization limited to the inner half of enamel up to dentin-enamel junction (2), demineralization limited to the outer half of dentin (3), or demineralization extending into the inner half of dentin (4). The straight-line tool was used to measure the depth of the lesion compared with the enamel or dentin thickness to aid in classification. Surfaces showing any disagreement between the 2 evaluators were reevaluated together until a consensus was reached for each surface. Table I shows the scoring system used for each evaluation method.

Table I

Comparison of severity measure for each evaluation method

Variables	Micro-CT	BWR	NIRI	Combination
Sound
No lesion present	0	0	0	0
Enamel Lesion			1	1
Lesion limited to outer half enamel	1	1
Lesion limited to inner half, not extending to DEJ	2	2
Dentin Lesion
Extending beyond DEJ but limited to outer half	3	3	2	2
Inner half	4	4

The extracted teeth were mounted into typodonts using Regisil VPS material (Dentsply Sirona, Charlotte, NC). Mounting sequences were blindly set up to simulate normal anatomy; premolars were never mounted between 2 molars. All surfaces of interest were mounted at adjacent sites to simulate interproximal contacts with a minimum of 3 teeth per quadrant. The mounted teeth continued to be stored in a sealed container at 4°C with DIW-soaked gauze covering the exposed surfaces.

Intraoral scans were carried out using iTero Element 5D with the NIRI scan feature turned on after a quick air dry to prevent water from obscuring the interproximal contacts. The teeth were submerged in DIW for approximately 24 hours before capturing the intraoral scan with a gentle air dry before scanning as early trials revealed better NIRI performance with hydrated teeth. The intraoral scan captured the entire mounted segment.

Using the My iTero web platform, the NIRI feature was hovered over each interproximal contact of interest. Contrast and brightness were adjusted to achieve optimal viewing for each surface. Two screenshots were captured for each surface of interest: the NIRI feature and the white light photograph. The images were assembled onto a slideshow for examiners to evaluate.

Two examiners (A.B.V. and A.A.R.) and dental specialists (operative and pediatric dentistry), each with >25 years of dental clinical experience but without significant NIRI interpretation experience, underwent two 45-minute in-person training sessions together regarding the use of NIRI. Training sessions included background information about NIRI technology, sample images and materials from the company’s clinical guide, and 8 sample surfaces captured from this study, displayed with the associated micro-CT classification. Because of the documented limitations in the ability of NIRI to precisely discriminate the depth of lesions, ^, evaluators were asked to use a modified version of the Söchtig classification, which combined the enamel categories and dentin categories: sound surface (0), carious lesion limited to enamel (1), and carious lesion extending into dentin (2). The 2 examiners thoroughly discussed the practice images, coming to a consensus on the description of interproximal caries. Each training session concluded with an independent evaluation of a set of practice teeth to assess calibration levels, followed by a discussion of disagreements.

Once both examiners were comfortable moving forward, the calibration sets were prepared. A slide was prepared for each surface, which presented the screenshot of the NIRI image next to the white light image, with an arrow indicating which surface to evaluate. A randomly selected set of 13 calibration surfaces, separate from the examination teeth, were evaluated independently by both examiners to assess interrater reliability (weighted kappa = 0.707; standard error = 0.16). After at least a 1-week interval, the examiners reevaluated the calibration set to assess intrarater reliability (examiner 1: weighted kappa = 1.00; examiner 2: weighted kappa = 0.860; standard error = 0.13).

After calibration, the examiners evaluated the full set of examination surfaces (n = 85). Surfaces were evaluated independently to assess interexaminer reliability between 2 trained and calibrated examiners.

BWRs were captured using the mounted samples to simulate the traditional radiographic examination. The FOCUS Instrumentarium x-ray unit (KaVo FOCUS, Charlotte, NC) was used to capture conventional BWRs using the manufacturer’s 6-cm-diameter collimator. Bitewings of each quadrant were taken multiple times with different angulations such that each contact surface of interest was fully opened. The radiographs were then converted into images at a standardized brightness and compiled onto standardized slides. For each surface of interest, the radiograph with the contact opened for examination was set up next to a clinical photograph. Evaluators were then presented with standardized slides to evaluate the areas of interest efficiently without manipulating the images.

The same 2 examiners who evaluated the NIRI images (A.B.V. and A.A.R.) were given a 90-day interval between the NIRI evaluation and the beginning of the BWR evaluation. Given the everyday use of BWR by our experienced cardiology examiners, we moved straight to the calibration. A set of 14 calibration surfaces, separate from the examination teeth, were randomly selected. A slide was prepared for each surface, which included the prepared BWR image and the white light image, using an arrow to indicate the surface of interest. Examiners were asked to use the following criteria: (0) sound surface, (1) lesion limited to the outer half of enamel, (2) lesion extends to the inner half of enamel but does not reach dentinoenamel junction (DEJ), (3) lesion extends beyond DEJ, but not to the inner half of dentin, and (4) lesion extends to the inner half of dentin. Interrater reliability for BWR evaluation was similar to literature values (weighted kappa = 0.449; standard error = 0.21).

After calibration, the examiners independently evaluated the full set of examination surfaces (n = 84).

Once again, the same 2 examiners (A.B.V. and A.A.R.) were given at least a 10-day interval between the previous evaluation and the beginning of the combination evaluation. Because of the previous calibration and the novelty of this exercise, no calibration was completed before the combination evaluation. A slide was created for each surface, which included the NIRI image, BWR image, and white light photograph, with arrows indicating the surface of interest. The examiners were asked to use the modified version of the Söchtig classification sound surface (0), carious lesion limited to enamel (1), and carious lesion extending into dentin (2). The examiners independently evaluated all surfaces (n = 98).

Statistical analysis

Examiners’ scores were input directly by examiners and stored in independent, blinded Excel (Microsoft, Redmond, Wash) spreadsheets. Data was compiled and exported to SAS (Cary, NC) for statistical analysis. Linear weighted Cohen’s kappa was used to evaluate interrater reliability and examiner agreement with micro-CT scoring. Interpretation of kappa scores followed by Landis and Koch : <0.00, no agreement; 0.00-0.20, slight agreement; 0.21-0.40, fair agreement; 0.41-0.60, moderate agreement; 0.61-0.80, substantial agreement; and 0.81-1.00, near perfect agreement. A paired chi-square test was used to evaluate differences between examiners. Agreement statistics included sensitivity, specificity, positive predictive value, negative predictive value, and area under the receiver operating characteristic. Stratification of reliability by lesion severity was found by making a table between 2 dummy variables of micro-CT reference standard and each examiner for each stratified level (micro-CT scores: 0, 1, and 2). The dummy variables were generated to determine whether the values (reference standard or each examiner) have a specific level.

Results

Micro-CT consensus reference standard classification of 98 interproximal surfaces revealed that 54.1% of surfaces were sound (n = 53), 18.4% had proximal lesions limited to the outer half of enamel (n = 18), 17.3% lesions limited to the inner half of enamel (n = 17), and 10.2% lesions extending into dentin, up to an outer third of dentin depth (n = 10). Interrater reliability between the 2 examiners for the full sample set is shown in Table II . For interrater reliability, NIRI alone showed moderate agreement, BWR alone showed poor agreement, and NIRI and BWR in combination showed fair agreement.

Table II

Only gold members can continue reading. Log In or Register to continue