In vitroperformance of near infrared light transillumination at 780-nm and digital radiography for detection of non-cavitated approximal caries

Abstract

Objectives

To evaluate the ability of a Near Infrared Light Transillumination (NILT) device to detect non-cavitated approximal caries lesions; and to compare its performance to Digital Radiography (DR).

Methods

Thirty human extracted premolars (sound to lesions into the outer one-third of dentin) were selected. Lesion depth was confirmed by micro-computed tomography (μ-CT). Teeth were mounted in a custom-made device to simulate approximal contact. DR and NILT (CariVu™, DEXIS, LLC, Hatfield, PA, USA) examinations were performed and repeated by three trained and calibrated examiners. Sensitivity, specificity, area under ROC curve (A z ), inter- and intra-class correlation coefficients (ICCs) for each method, and correlation among the methods were determined.

Results

ICCs for intra-/inter-examiner agreement were substantial for NILT (0.69/0.64), and moderate for DR (0.52/0.48). Sensitivity/specificity for NILT and DR were 0.68/0.93 and 0.50/0.64, respectively. A z for NILT was 0.81, while for DR it was 0.61. Spearman correlation coefficient with μ -CT for NILT (0.65, p < 0.001) demonstrated moderate association, while that of DR suggested no association (0.19, p = 0.289).

Conclusion

Within the limitations of this in vitro study, NILT demonstrated a potential for early approximal caries detection. NILT and DR performed the same regarding the accuracy for non-cavitated approximal caries detection; however, NILT was superior to DR in terms of repeatability, agreement and correlation with μ -CT.

Clinical significance

A commercial version of NILT was recently introduced as a non-irradiative adjunctive caries detection method. It uses near infrared (NIR) light at 780-nm to transilluminate teeth and captures live images from the occlusal surface. This study demonstrates that NILT can be used as an alternative to radiography for non-cavitated approximal caries detection.

Introduction

It has been known that demineralization and remineralization of tooth structure occur over time. More importantly, the net balance between the pathologic and protective factors will determine the rate of caries lesion progression to a state that can be detected visually or by other caries detection methodologies . Because of the widespread use and availability of fluoride, caries lesion behavior has changed significantly and a lower progression rate has been observed . The slow progress of the caries lesion gives the dental professional a substantial opportunity to diagnose and manage dental caries at an early stage, as non-cavitated lesions can be arrested or remineralized before irreversible destruction of the tooth structure occurs. Hence, early caries detection and monitoring can have a profound effect on the success of preventive treatment of non-cavitated caries lesions .

First described by Raper in 1925 , bitewing radiography in combination with visual examination has become the traditional method for approximal caries detection . Radiographs also help to estimate approximal and occlusal caries lesion depth and enable detection of lesions on visually inaccessible surfaces. A major limitation of radiographic caries detection is that image interpretation may vary significantly within or between examiners . Furthermore, two recent systematic reviews with meta-analysis evaluated the performance of visual examination and radiography for caries lesion detection . They demonstrated that these two conventional methods have low sensitivity but high specificity for early approximal caries detection. This means that the conventional methods have high risk to miss detection of approximal lesions.

An argument has been raised to re-evaluate clinicians’ over-reliance on using radiography for caries detection because of the following reasons: 1) decrease in caries prevalence; 2) the risk associated with low-dose radiation, especially for children ; and 3) slow progression rate of approximal caries lesions as the result of the widespread use of fluoride . Although there is no conclusive evidence that dental radiographs taken during childhood increase the risk of malignant diseases , it is still difficult to justify the repeated use of bitewing radiography to monitor lesions and consequently to evaluate the effectiveness of non-invasive dental preventive treatments . Therefore, because of these concerns, it is of vital importance to continue the development of new caries detection techniques with improved sensitivity, specificity, and reliability, while decreasing risk to the patients.

The search for a highly accurate caries detection method has resulted in many attempts. Methods based on visible and near infrared (NIR) light transillumination have been under development since the early 1990s . Thereafter, Digital Imaging Fiber Optic Transillumination (DIFOTI) (Electro-Optical Sciences Inc., NY, USA) was introduced as a more sensitive, non-irradiative adjunctive method for early caries detection . Its priciple is based on transilluminating the tooth with a visible light (wavelength between 400-nm to 700-nm) and the caries lesion will appear as a dark shadow due to differences in light scattering and absorption of light photons. However, visible light cannot penetrate deep into enamel because of high light scattering by tooth tissues. NIR light can penetrate deeper into enamel due to reduced scattering in tooth enamel . NIR transillumination for caries detection has been under further development since early last decade . NIR light at 1310-nm has shown an enhanced image contrast over 830-nm for approximal caries transillumination. However, both have shown highly improved image contrast compared to wavelengths in the visible range (400-nm to 700-nm) . Furthermore, two in vivo studies evaluated a prototype of NIR transillumination . Staninec et al., illustrated that lesions appearing in radiographs were also detected in NIR transillumination . Simon et al. investigated the performance of NIR transillumination and radiography clinically on premolars planned for extraction with microradiography as a gold standard. The NIR transillumination showed higher sensitivity than radiography for approximal caries detection .

A commercial near infrared light transillumination (NILT) device was introduced in Europe as DIAGNOcam (KaVo, Biberach, Germany) in 2012, and a year later, in the United States as CariVu™ (DEXIS, LLC, Hatfield, PA, USA) . This device uses a near infrared (wave length ∼780-nm) light to transilluminate the tooth. This system consists of a charged coupled device (CCD) sensor to capture images, connection to a computer, special software, and elastic arms containing a NIR light source that transmits light through the gingiva, the alveolar bone, the root of the tooth, and up to the crown. The image displays and saves from the occlusal surface . To our best knowledge, no study has evaluated and compared the commercially available NILT to a histological reference method that can identify the severity of caries lesions. Hence, there is a need for further evaluation of NILT regarding the detection of early non-cavitated caries lesions. For practical reasons, we designed an in vitro model to simulate the clinical situation as close as possible in terms of allowing the near infrared light to be transmitted through a certain material thickness, the roots and up to the crown of the tooth. We employed non-destructive micro-computed tomography ( μ -CT) as the histological reference method to confirm lesion depth extension. The objectives were to evaluate the ability of NILT to detect non-cavitated approximal caries lesions; and to compare between NILT and digital radiography (DR). The null hypothesis was there was no a difference in the ability of caries detection between NILT and DR.

Materials and methods

Teeth selection

Eighty-five sound and carious human premolars were selected from a pool of extracted teeth. The non-cavitated caries lesions were located on approximal surfaces and they were surrounded by sound enamel. The presence of caries was determined by visual tooth surface changes . The extracted human teeth were collected from dental practitioners across the United States and transported in 0.1% thymol solution. The collection of human teeth for use in dental laboratory research studies has been approved by the Indiana University (IU) Institutional Review Board. All specimens were kept in 0.1% thymol solution at 4°C until used. Teeth were cleaned using Robinson’s brush with water on a slow speed handpiece.

Initial microfocus computed tomography (μ-CT) image acquisition

The teeth were mounted and secured on plastic Lego ® bricks (The LEGO Group, Billund, Denmark) using utility wax (Heraeus Kulzer Inc., Lafayette, IN, USA). The teeth were scanned using microfocus computed tomography ( μ -CT) for lesion depth determination and to establish a gold standard assessment. The μ -CT images were acquired using Skyscan μ -CT machine (Skyscan 1172, Kontich, Belgium) at 80 kV, 134 μ A, 8.9 μm pixel size resolution. An Al + Cu filter was used. The specimens were rotated at 180° with rotation step of 0.7° and frame average of 4. Two-dimensional image reconstructions were done using NRecon version 1.6.6 software (Bruker microCT, Kontich, Belgium) and stored in 16-bit TIFF files. Visual interpretation of sagittal views of the μ -CT images was performed using image display software (CT-Analyzer, Bruker microCT, Kontich, Belgium) in a dark room on a digital screen (DELL, U2412Mb, Limerick, Ireland). The images were evaluated by two examiners (NA, MA) according to the criteria previously prescribed as described in Table 1 . For each specimen, the image with the deepest lesion was considered for score determination. In case of disagreement, the examination was performed again until consensus agreement was achieved.

Model assembly

Carious teeth with lesion extension into the inner two-thirds of dentin (D 2 : Score 4 lesions, Table 1 ) were excluded. Also, cracked teeth and teeth with obvious fluorosis were excluded. Eventually, twelve extracted premolars were selected based on lesion depth extension according to μ -CT for training and calibration, and thirty teeth for the main examination. The sample distribution for training and calibration/main examination was as follows: sound surface [E 0 (Score 0): n = 3/12]; lesion in the outer half of the enamel [E 1 (Score 1): n = 3/6]; lesion in the inner half of the enamel [E 2 (Score 2): n = 3/6]; lesion in the outer one-third of the dentin [D 1 (Score 3): n = 3/6], respectively.

The teeth were mounted on plastic Lego ® bricks (The LEGO Group, Billund, Denmark) with the test surface adjacent to a sound tooth. The height of contact, lesion and marginal ridge were standardized at the same level for all specimens. Triad ® visible light cure resin (DENTSPLY International, Inc., York, PA, USA) was applied around the root and the cervical part of the teeth at the level of the cemento-enamel junction resembling the anatomy of the gingiva. The selection of Triad ® visible light cure resin was based on a pilot study. Several extracted premolars were mounted using different imbedding materials. NILT images were acquired and evaluated in order to obtain images with contrast comparable with in vivo NILT images available in public sources . Dental floss was used to confirm the presence of the proximal contact ( Fig. 1 ). The assembled models were kept in a sealed plastic container with wet gauze to maintain humidity. After model assembly, the mounted teeth (specimens) were again scanned using μ -CT with the same setting as described previously to standardize μ -CT image quality.

Fig. 1
Sample assembly on Lego ® bricks. a. Adjacent sound tooth; b. Examined/Test tooth.

Digital radiography (DR) image acquisition

The mounted teeth were placed on a custom film holder with a beam-aiming device. The DR images were obtained using Schick 33 CDR sensor (Sirona Dental Inc., Long Island City, NY, USA) at 60 kV, 7 mA/0.20 s (Sirona, Heliodent DS. Bensheim, Germany). Plexiglass, 10 × 2.5 × 10 cm, was placed to simulate the soft tissue effect between the x-ray tube head and the tooth. The images were saved using dedicated software (CDR ® DICOM, Schick Technologies Inc., Long Island City, NY, USA). The images were extracted and stored in uncompressed TIFF format.

Training and calibration

Prior to the main examination, three examiners (MA, AH, AG), who had more than 10 years of clinical teaching and research experience, were trained on DR and NILT examinations. The training course included theoretical elements in a PowerPoint presentation for one hour, and hands-on training on previously obtained specimens for NILT and DR images for three hours.

The primary and repeated calibration examinations were performed during separate sessions. One examiner (NA) randomly ordered the samples between examiners and before each examination for each method using the random function of Microsoft Excel software (Microsoft ® Excel ® version 14.6.0, Microsoft Corporation, Redmond, WA, USA).

DR calibration

The images were displayed randomly on a digital screen in a dark room (DELL, U2412Mb, Limerick, Ireland) via image viewer software (Windows Photo Viewer, version Windows 7, Microsoft, Redmond, WA, USA). The images were evaluated by the three examiners according to the criteria previously prescribed in Table 1 .

NILT calibration

After air-drying, each specimen was examined with NILT (CariVu™, DEXIS, LLC, Hatfield, PA, USA) in a dark room. The examiners were instructed to place the light aperture and make it contact the resin that resembled the gingiva; the NILT camera was centered perpendicularly over the test tooth. The live picture was monitored on the digital screen (DELL, U2412Mb, Limerick Ireland) and when the examiners were satisfied with a viewed image, they were asked to capture it and report the score. The images were saved through the software (DEXIS, version 9.4.0, Hatfield, PA, USA). The scoring criteria are described in Table 1 .

Repeated calibration

Each examiner performed DR and NILT calibration again on the same specimens at least two days after the initial calibration and in the same manner as previously described.

Statistical analysis for calibration

Intra-examiner repeatability and inter-examiner agreement of all methods were assessed using intraclass correlation coefficients (ICCs). Two-way tables were also used to provide additional information about the repeatability and agreement.

Calibration results

The intra-examiner repeatability ICCs after calibration were as follows: DR (0.86) and NILT (0.71). The inter-examiner agreement ICCs were as follows: DR (0.86) and NILT (0.71). Since the ICCs values was not satisfactory for NILT a second training and calibration exercise was performed only for NILT as previously described. The intra-examiner repeatability ICC for the second NILT calibration was 0.81, and the inter-examiner agreement ICC was 0.81.

Main examinations

Main DR and NILT examinations were performed on the thirty teeth (n = 30) as previously prescribed in “model assembly.” It was performed in the same manner as in calibration for each method. Repeated examination was performed one week after the first main examination on all teeth (n = 30).

Sample size justification

Data from previous studies indicated a correlation of approximately 0.7 between methods. With a sample size of 10 sound teeth and 5 teeth for each of E 1 , E 2 , and D 1 , the study had 80% power to detect a difference in the area under Receiver Operating Characteristic (ROC) curve (A Z ) of 0.23 (0.67 vs. 0.90), assuming a two-sided test with a 5% significance level.

Statistical analysis

Intra-examiner repeatability and inter-examiner agreement of all methods were assessed using intraclass correlation coefficients (ICCs). Two-way tables were also used to provide additional information about the repeatability and agreement. Comparisons between the DR and NILT methods for sensitivity, specificity, and A Z were performed using bootstrap analyses. The bootstrap analysis method employed in this study allowed us to use the individual data from all three examiners, which is more clinically relevant than using a rule to combine scores across multiple examiners, such as using the maximum score, majority rule, or forced consensus. The bootstrap methodology uses resampling techniques to estimate statistics and perform comparisons for values that are not normally distributed. In this case, the bootstrap also provides a way to properly account for the correlations between examiners, between repeats, and between methods when all were assessed on the same sample. The sensitivity was determined further based on three μ -CT thresholds: lesion in the outer half of the enamel [E 1 ( μ -CT = E 1 )]; lesion in the inner half of the enamel [E 2 ( μ -CT = E 2 )]; and lesion in the outer one-third of the dentin [D 1 ( μ -CT = D 1 )]. The correlations among the measurements and the correlations of the measurements with the μ -CT were also calculated using bootstrap methods.

Materials and methods

Teeth selection

Eighty-five sound and carious human premolars were selected from a pool of extracted teeth. The non-cavitated caries lesions were located on approximal surfaces and they were surrounded by sound enamel. The presence of caries was determined by visual tooth surface changes . The extracted human teeth were collected from dental practitioners across the United States and transported in 0.1% thymol solution. The collection of human teeth for use in dental laboratory research studies has been approved by the Indiana University (IU) Institutional Review Board. All specimens were kept in 0.1% thymol solution at 4°C until used. Teeth were cleaned using Robinson’s brush with water on a slow speed handpiece.

Initial microfocus computed tomography (μ-CT) image acquisition

The teeth were mounted and secured on plastic Lego ® bricks (The LEGO Group, Billund, Denmark) using utility wax (Heraeus Kulzer Inc., Lafayette, IN, USA). The teeth were scanned using microfocus computed tomography ( μ -CT) for lesion depth determination and to establish a gold standard assessment. The μ -CT images were acquired using Skyscan μ -CT machine (Skyscan 1172, Kontich, Belgium) at 80 kV, 134 μ A, 8.9 μm pixel size resolution. An Al + Cu filter was used. The specimens were rotated at 180° with rotation step of 0.7° and frame average of 4. Two-dimensional image reconstructions were done using NRecon version 1.6.6 software (Bruker microCT, Kontich, Belgium) and stored in 16-bit TIFF files. Visual interpretation of sagittal views of the μ -CT images was performed using image display software (CT-Analyzer, Bruker microCT, Kontich, Belgium) in a dark room on a digital screen (DELL, U2412Mb, Limerick, Ireland). The images were evaluated by two examiners (NA, MA) according to the criteria previously prescribed as described in Table 1 . For each specimen, the image with the deepest lesion was considered for score determination. In case of disagreement, the examination was performed again until consensus agreement was achieved.

Model assembly

Carious teeth with lesion extension into the inner two-thirds of dentin (D 2 : Score 4 lesions, Table 1 ) were excluded. Also, cracked teeth and teeth with obvious fluorosis were excluded. Eventually, twelve extracted premolars were selected based on lesion depth extension according to μ -CT for training and calibration, and thirty teeth for the main examination. The sample distribution for training and calibration/main examination was as follows: sound surface [E 0 (Score 0): n = 3/12]; lesion in the outer half of the enamel [E 1 (Score 1): n = 3/6]; lesion in the inner half of the enamel [E 2 (Score 2): n = 3/6]; lesion in the outer one-third of the dentin [D 1 (Score 3): n = 3/6], respectively.

The teeth were mounted on plastic Lego ® bricks (The LEGO Group, Billund, Denmark) with the test surface adjacent to a sound tooth. The height of contact, lesion and marginal ridge were standardized at the same level for all specimens. Triad ® visible light cure resin (DENTSPLY International, Inc., York, PA, USA) was applied around the root and the cervical part of the teeth at the level of the cemento-enamel junction resembling the anatomy of the gingiva. The selection of Triad ® visible light cure resin was based on a pilot study. Several extracted premolars were mounted using different imbedding materials. NILT images were acquired and evaluated in order to obtain images with contrast comparable with in vivo NILT images available in public sources . Dental floss was used to confirm the presence of the proximal contact ( Fig. 1 ). The assembled models were kept in a sealed plastic container with wet gauze to maintain humidity. After model assembly, the mounted teeth (specimens) were again scanned using μ -CT with the same setting as described previously to standardize μ -CT image quality.

Jun 17, 2018 | Posted by in General Dentistry | Comments Off on In vitroperformance of near infrared light transillumination at 780-nm and digital radiography for detection of non-cavitated approximal caries

VIDEdental - Online dental courses

Get VIDEdental app for watching clinical videos