Comparative assessment of occlusal and proximal caries detection using an intraoral scanner and panoramic radiography

Abstract

Background

Recently, fluorescence-based caries detection has been integrated into intraoral scanners as a radiation-free adjunctive tool for caries assessment. However, evidence regarding the agreement of these systems with conventional diagnostic methods remains limited. Therefore, the aim of this retrospective study was to evaluate the agreement among panoramic radiography, the fluorescence-based caries detection mode of the TRIOS 5 intraoral scanner, and ICDAS (International Caries Detection and Assessment System)-based on-screen assessment of digital intraoral scan images in the detection and classification of occlusal and approximal carious lesions.

Methods

Sixty patients aged 16–45 years who met the inclusion criteria were included. Archived panoramic radiographs and TRIOS 5 (3Shape TRIOS A/S, Copenhagen, Denmark) intraoral scan records were retrospectively evaluated. Permanent teeth were analyzed. Panoramic radiographic findings were classified as Score 0 (sound), Score 1 (initial lesion), and Score 2 (deep dentin lesion). The fluorescence-based caries detection mode of the TRIOS 5 classified lesions as initial, moderate/extensive, or insufficient scan. Archived intraoral scan images were also visually assessed using an on-screen ICDAS-based protocol. Inter-method agreement was evaluated using Gwet’s AC1/AC2, Fleiss’ kappa, and Krippendorff’s alpha coefficients. McNemar’s test and complementary mixed-effects logistic regression were used for pairwise comparisons. Sensitivity, specificity, predictive values, likelihood ratios, and diagnostic odds ratios were also calculated. Statistical significance was set at p < 0.05.

Results

The fluorescence-based caries detection mode of the TRIOS 5 classified a higher proportion of teeth as having occlusal and enamel-limited lesions, whereas panoramic radiography classified a higher proportion as having approximal and deep dentin lesions. Agreement between TRIOS 5 and the on-screen ICDAS-based assessment was higher than that between TRIOS 5 and panoramic radiography, although overall agreement among the three methods remained low. In subgroup analyses, agreement between TRIOS 5 and the visual assessment was higher for molars than for premolars.

Conclusions

The fluorescence-based caries detection mode of the TRIOS 5 intraoral scanner provided diagnostic information complementary to panoramic radiography. Further prospective studies using independent reference standards are required to confirm its clinical performance.

Introduction

Despite increasing awareness of oral health, dental caries remains one of the most prevalent chronic diseases worldwide. According to the World Health Organization, 60–90% of school-aged children and nearly all adults in developing countries are affected by dental caries [1]. Therefore, one of the principal goals of contemporary dentistry is the early detection of carious lesions before irreversible hard tissue destruction, pain, and tooth loss occur [2].

Occlusal surfaces are highly susceptible to caries development because of their complex pit and fissure morphology, which facilitates plaque retention. Consequently, initial enamel lesions may develop on apparently intact surfaces and often remain undetected until they progress into dentin [2]. Since occlusal lesions generally become radiographically visible only after dentinal involvement, clinicians frequently rely on visual changes such as discoloration, loss of translucency, and tactile examination with dental explorers [3].

In contrast, approximal surfaces present greater diagnostic difficulties because they cannot be directly visualized due to the presence of adjacent teeth. For this reason, bitewing radiography has traditionally been considered the standard adjunctive imaging method for proximal caries detection [4]. Several studies and systematic reviews have demonstrated that bitewing radiographs provide relatively high specificity for proximal caries detection; however, their sensitivity for early enamel lesions remains limited [5, 6].

Although bitewing radiographs are widely recommended for proximal caries diagnosis, their routine use in large clinical settings may be limited by several practical challenges. Intraoral sensor placement may induce discomfort, gag reflex, and nausea in some patients, particularly pediatric patients and individuals with strong gag reflex sensitivity [7]. Additionally, obtaining bitewing radiographs may increase chairside time in high-volume university clinics and public dental institutions with substantial patient flow. Consequently, panoramic radiography remains one of the most frequently obtained radiographic examinations in dental faculties and large outpatient clinics because it provides comprehensive diagnostic information, is routinely included in initial patient examinations, and is commonly archived in institutional databases [8, 9].

From a research perspective, panoramic radiographs may offer an important advantage in retrospective studies because they are significantly easier to access through institutional archives compared with bitewing radiographs, which are not routinely available for every patient. Therefore, retrospective evaluations based on panoramic radiography may provide a more practical and accessible diagnostic dataset in large patient populations [8].

To overcome the limitations of conventional diagnostic approaches, several non-invasive technologies have been developed over the past decades, including Quantitative Light-Induced Fluorescence (QLF), DIAGNOdent, Fiber Optic Transillumination (FOTI), Digital Imaging Fiber Optic Transillumination (DIFOTI), Optical Coherence Tomography (OCT), laser fluorescence, electrical conductance systems, Near-Infrared Imaging (NIRI) and fluorescence-based imaging technologies [10].

Among these technologies, fluorescence-based systems using blue-violet light have attracted increasing attention because they enable radiation-free detection of early enamel lesions [11]. More recently, this technology has been incorporated into intraoral scanners (IOS), allowing simultaneous three-dimensional (3D) digital acquisition and fluorescence-based caries assessment during a single scanning procedure. In addition to producing detailed digital models of oral structures, these systems are increasingly being investigated for clinical applications such as caries detection, lesion monitoring, and treatment documentation [12].

One of the latest systems in this field is TRIOS 5 (3Shape TRIOS A/S, Copenhagen, Denmark), which combines three-dimensional intraoral scanning with fluorescence-based caries detection. The system uses blue-violet light at a wavelength of 405 nm to induce fluorescence from tooth structures and bacterial metabolites, while the emitted fluorescence at wavelengths greater than 435 nm is captured and analyzed to generate a color-coded diagnostic overlay indicating the presence and severity of carious lesions [13]. This integrated approach enables simultaneous acquisition of a three-dimensional digital model and fluorescence-based caries assessment within a single scanning procedure.

In addition to radiographic and fluorescence-based examination, visual caries assessment remains an essential diagnostic component. To improve reproducibility and standardization in visual examinations, the International Caries Detection and Assessment System (ICDAS) was introduced as a validated scoring method for classifying carious lesions according to their severity, ranging from initial enamel changes to extensive cavitation [14]. Previous studies have demonstrated that ICDAS can be applied to the visual assessment of digital intraoral scan images, with on-screen evaluation showing diagnostic performance comparable to that of conventional visual examination [15, 16]. Accordingly, in the present study, three-dimensional digital intraoral scan images were visually assessed using the on-screen ICDAS criteria and used as the third method for comparison with panoramic radiography and the TRIOS 5 caries detection aid mode.

Recent clinical studies have increasingly explored the use of fluorescence-based intraoral scanners for caries detection in permanent dentition, highlighting their potential as adjunctive diagnostic tools [17, 18]. Nevertheless, important gaps remain in the current literature. Most previous studies have focused on either occlusal or approximal lesions and have primarily compared intraoral scanner–based caries detection with bitewing radiography. Moreover, published studies specifically evaluating the caries detection performance of the TRIOS 5 system remain scarce. Furthermore, little is known about the agreement between archived panoramic radiographs, TRIOS 5 caries detection aid, and on-screen assessment of the corresponding three-dimensional digital scan images within the same patient population. Evaluating the agreement among these routinely available diagnostic records may provide valuable information for retrospective clinical studies, support more efficient use of existing diagnostic data, and potentially contribute to the earlier identification of carious lesions without requiring additional radiographic examinations.

Therefore, the aim of the present retrospective study was to compare caries detection outcomes obtained from archived panoramic radiographs, the TRIOS 5 caries detection mode, and on-screen ICDAS-based assessment of three-dimensional intraoral scan images within the same patient population. The primary research question was whether these three diagnostic approaches demonstrated comparable agreement in the detection of dental caries. It was hypothesized that the diagnostic findings obtained using the TRIOS 5 caries detection mode would demonstrate concordance with those obtained from panoramic radiography and visual assessment of three-dimensional intraoral scan images in terms of caries detection, lesion type, and lesion severity.

Materials and methods

The study included patients who presented for routine dental treatment and had previously undergone panoramic radiographic imaging together with intraoral scanning as part of standard clinical care. After ethical approval was obtained (Approval No: 2025 − 233), archived patient records were retrospectively screened through the institutional digital patient information system.

Patient selection

Patient records were retrospectively identified through the institutional digital database following ethical approval. Records of patients treated between April 14, 2025, and December 31, 2025 were screened for eligibility. A total of 90 patient records were initially assessed.

Individuals aged 16–45 years with permanent dentition were considered eligible. Patients were included if archived panoramic radiographs obtained within the previous three months and corresponding TRIOS 5 intraoral scan records of sufficient diagnostic quality were available.

After application of the predefined inclusion and exclusion criteria, 60 patients (39 females and 21 males) were included in the final analysis. A total of 511 eligible permanent teeth were evaluated according to the predefined study protocol. Each tooth was classified according to the predominant caries location (occlusal or approximal) and included as a single observation in the statistical analyses.

Inclusion and exclusion criteria

To minimize potential sources of diagnostic bias and ensure standardized evaluation across all diagnostic methods, records with poor-quality panoramic radiographs, incomplete intraoral scan records, restorations interfering with caries detection, fixed orthodontic appliances, developmental enamel defects, third molars, or any other condition that could affect diagnostic interpretation were excluded. Only teeth that were fully visible on both the panoramic radiograph and the corresponding intraoral scan were included in the analyses.

Sample size calculation

The minimum required sample size was estimated using the formula proposed by Buderer [19], which is specifically designed for diagnostic performance studies by incorporating the expected sensitivity, specificity, disease prevalence, and desired precision. Although all eligible archived records within the study period were retrospectively screened, a priori sample size estimation was performed to verify that the available sample would provide adequate statistical precision for the planned analyses. Calculations were conducted using a 95% confidence level (α = 0.05; Z = 1.96), an expected sensitivity of 0.90, specificity of 0.90, disease prevalence of 0.30, and an acceptable margin of error of ± 0.05. Accordingly, the minimum required sample size was calculated as 461 observations based on sensitivity and 198 based on specificity; therefore, the larger estimate (n = 461) was adopted. Allowing for a potential 10% data loss, the target sample size was increased to 507 observations. Following the retrospective screening, 511 eligible teeth with corresponding panoramic radiographs and TRIOS 5 intraoral scan records were included in the final analyses.

Examiner blinding and inter-examiner agreement

Panoramic radiographs and visual assessments of the three-dimensional (3D) digital intraoral scan models were independently evaluated by two examiners (N.A., Associate Professor, and F.Ö.T., postgraduate resident in Restorative Dentistry). Both examiners were experienced in restorative dentistry and familiar with the predefined scoring systems applied in the present study. Each examiner recorded the findings independently without any discussion, guidance, or consensus throughout the assessment process. Inter-examiner reliability for the presence of caries was therefore assessed across the entire study dataset rather than in a subset of observations.

To minimize observer and recall bias, the three diagnostic modalities were evaluated in separate assessment sessions. Panoramic radiographs were evaluated first, followed by on-screen ICDAS-based visual assessment of the three-dimensional (3D) intraoral scan models after a 15-day interval, and finally by evaluation of the same scans with the software-integrated Caries Detection Aid enabled after an additional 15-day interval. During each assessment session, the examiners were blinded to the findings of the other diagnostic methods, to each other’s evaluations, and to all patient-identifying information.

Agreement between the two examiners was evaluated using Cohen’s kappa coefficient. The kappa values were 0.966 for panoramic radiographic evaluations and 0.978 for visual assessments of the 3D digital scan models, indicating almost perfect inter-examiner agreement. These findings indicate a high level of consistency between the two evaluators and minimal observer-related variability. Accordingly, the ratings of the primary examiner were used for the subsequent statistical analyses.

Panoramic radiographic evaluation

Archived panoramic radiographs obtained from the Department of Oral Diagnosis and Radiology were retrospectively analyzed for the detection of occlusal and approximal carious lesions. All radiographs had been acquired as part of routine clinical examinations using the same institutional digital panoramic imaging system (Planmeca ProMax^® 3D panoramic imaging system, Planmeca Oy, 00880 Helsinki, Finland) operating at 66 kVp and 8 mA, with an exposure time of 15.8 s. The images were retrieved from the institutional Picture Archiving and Communication System (PACS). Image evaluations were performed on the same diagnostic monitor under standardized viewing conditions with natural ambient lighting.

During image interpretation, image magnification as well as brightness and contrast adjustments were permitted when necessary to optimize visualization. Only permanent teeth were included in the radiographic evaluation, and each eligible tooth was independently evaluated using a three-category radiographic scoring system. Radiographic findings were classified as Score 0 (sound), Score 1 (initial lesion), and Score 2 (deep dentin lesion).

Intraoral scanner evaluation

The scanning protocol was carried out by trained restorative dentistry specialists who had received formal training in the use of the TRIOS 5 (3Shape TRIOS A/S, Copenhagen, Denmark) intraoral scanner equipped with the L1P-1 F scanner tip. As part of the routine clinical protocol, all patients had undergone professional dental prophylaxis within the previous six months and were instructed to brush their teeth immediately before scanning. During image acquisition, the dental operating light was switched off, external light sources were minimized, and the tooth surfaces were clean and dry. The manufacturer’s recommended scanning strategy was followed throughout image acquisition.

For the present study, the archived three-dimensional digital datasets were retrospectively accessed through the TRIOS Patient Monitoring (version 18.1; 3Shape A/S, Copenhagen, Denmark) system on 15-inch laptop monitors under standardized viewing conditions with natural ambient lighting. Each scan was evaluated using the software-integrated Caries Detection Aid (Fig. 1) enabled, which displays a fluorescence-based color overlay superimposed on the three-dimensional digital model to assist in the identification of carious lesions.

Fig. 1.

Automatic classification of lesions, with yellow overlays indicating initial lesions, red overlays indicating moderate/extensive lesions, and white overlays indicating insufficient scans when inadequate tooth color and/or fluorescence data prevented automatic caries score calculation

According to the software, tooth surfaces without fluorescence changes were displayed without color and interpreted as sound surfaces, whereas a yellow color indicated initial enamel-dentin caries lesions, and a red color indicated deep dentin caries lesions. When insufficient tooth color and/or fluorescence information prevented automatic caries detection, the corresponding teeth were classified by the software as insufficient scans.

For comparison with the other diagnostic methods, the software-generated color classifications were directly converted into numerical scores as follows: Score 0 (no carious lesion detected by the software), Score 1 (yellow; initial lesion), and Score 2 (red; moderate/extensive lesion). These numerical codes were recorded without modification and used in all statistical analyses.

Visual assessment of 3D digital intraoral scans

In addition to the automated caries detection analysis, standard three-dimensional (3D) intraoral scan images (Fig. 2) were independently evaluated by the examiners. The visual assessment was included to compare the clinician’s interpretation with the fluorescence-based output generated from the same 3D intraoral scan dataset, rather than to serve as an independent reference standard.

Fig. 2.

Three-dimensional image of the model obtained from the intraoral scan. The reference image used when performing an on-screen assessment according to ICDAS without the caries detection aid mode of the TRIOS 5 software being activated

The assessments were performed using the same TRIOS Patient Monitoring software on 15-inch laptop monitors under standardized viewing conditions with natural ambient lighting. The 3D models could be freely rotated, magnified, and repositioned to allow optimal visualization of all tooth surfaces.

The International Caries Detection and Assessment System (ICDAS) is a widely accepted scoring system for detecting and classifying dental caries severity [14]. To facilitate standardized on-screen assessment of digital models and reduce observer subjectivity, a previously validated modified ICDAS index incorporating fluorescence features was used for the visual evaluation [20]. A standardized on-screen visual assessment protocol guided all examinations. The evaluated teeth were classified according to ICDAS criteria as follows: Code 0 indicated sound surfaces with no visible caries. Codes 1–2 represented initial enamel changes without cavitation or dentinal involvement. Code 3 indicated localized enamel breakdown without visible dentin exposure. Code 4 represented lesions with underlying dentinal shadowing or visible dentin discoloration. Codes 5–6 indicated extensive caries with distinct cavities and visible dentin.

To facilitate comparison with the other diagnostic methods, enamel-limited lesions (ICDAS Codes 1–3) were grouped as Score 1, whereas dentin-involved lesions (ICDAS Codes 4–6) were grouped as Score 2 for the statistical analyses.

Statistical analysis

In the present study, three different assessment methods—panoramic radiography, the caries detection mode of the TRIOS 5 intraoral scanner, and visual evaluation of 3D digital scan data—were used on the same sample of teeth to evaluate inter-method agreement and comparative diagnostic performance in terms of caries presence (present/absent), caries type (occlusal/approximal), and lesion depth (enamel-limited/dentin-involved). Analyses were performed both for all teeth collectively and separately for molar and premolar subgroups. Due to the low prevalence of caries in anterior teeth, only descriptive statistics were used for this group.

Descriptive data were presented as frequency (n) and percentage (%). In the analysis of inter-method agreement, Gwet’s AC coefficients, which are considered more robust against the “prevalence paradox” were preferred over Kappa statistics due to prevalence imbalance within the dataset. Gwet’s AC1 (Agreement Coefficient 1) was used for analyses of caries presence and caries type, whereas weighted Gwet’s AC2 (Agreement Coefficient 2) was applied for analyses of lesion depth, which exhibited an ordinal structure. Fleiss’ Kappa coefficients were used for multi-method agreement analyses, and Krippendorff’s alpha (α) coefficients were applied for ordinal data. For all agreement coefficients and diagnostic performance measures, 95% confidence intervals were estimated using a patient-level clustered bootstrap procedure with 1,000 bootstrap replicates.

In addition, the diagnostic performance of TRIOS 5 was analyzed using panoramic radiography and visual assessment as reference methods by calculating sensitivity, specificity, positive and negative predictive values (PPV/NPV), likelihood ratios (LR+/LR−), and diagnostic odds ratio (DOR) metrics. Because multiple teeth were evaluated within the same participant, complementary mixed-effects logistic regression models were fitted with patient included as a random intercept to account for within-patient clustering. These models were used to compare the probability of a positive classification between methods while preserving the hierarchical structure of the data. McNemar’s test was used only to compare paired marginal positive classification rates between methods and was not interpreted as a measure of agreement or diagnostic performance.

Following the initial analyses, the integrity of the analytical dataset was independently verified by rechecking tooth identifiers, jaw and side orientation, surface matching, and outcome coding across all diagnostic methods. Statistical significance was set at α = 0.05. All analyses were performed using IBM SPSS Statistics v24.0, Jamovi v2.4, Microsoft Excel 2016, and RStudio v2023.09.0 software.

Given the retrospective design of the study and the use of routinely archived clinical records, diagnostic performance parameters were calculated based on comparisons between the findings obtained from the TRIOS 5 Caries Detection Aid and the available comparator methods. These parameters were interpreted in conjunction with inter-method agreement analyses to provide a more comprehensive evaluation of the relationship between the diagnostic approaches.

Results

Overall, agreement among the three diagnostic methods for caries detection was low. Although TRIOS 5 classified a greater proportion of teeth as carious than panoramic radiography and visual assessment, the highest pairwise agreement was observed between TRIOS 5 and the corresponding on-screen visual assessment, whereas agreement with panoramic radiography remained low (Tables 1 and 2).

Table 1.

Caries detection according to the assessment methods

Caries Detection	Panoramic Radiography	TRIOS 5	3D digital scan images
	n (%)	n (%)	n (%)
No caries	240 (47,0)	202 (39,5)	302 (59,1)
Caries present	271 (53,0)	309 (60,5)	209 (40,9)
Total	511 (100,0)	511 (100,0)	511 (100,0)

Table 2.

Agreement levels between assessment methods for caries detection

Source	n	Observed Agreement (Po)	Agreement Coefficient	%95 CI	p
Panoramic Radiography - TRIOS 5	511	%24,9	-0,476G	-0,579 - -0,366	< 0,001
Panoramic Radiography − 3D digital scan images	511	%37,0	-0,256G	-0,374 - -0,145	< 0,001
3D digital scan images - TRIOS 5	511	%62,0	0,241G	0,136–0,352	< 0,001
Overall Agreement (Panoramic – TRIOS 5–3D Digital Scan Images)	511	%11,9	-0,175F	-0,222 - -0,130	< 0,001

G Gwet’s AC1, F Fleiss’ Kappa, CI Confidence interval. Confidence intervals were obtained by a 1000-repetition patient-level clustered bootstrap percentile method

When panoramic radiography was used as the comparator, agreement between panoramic radiography and the TRIOS 5 caries detection mode was limited. Only a minority of teeth (approximately one-third) classified as carious on panoramic radiographs received the same classification by TRIOS 5 (Fig. 3), whereas a substantial proportion of teeth classified as sound on panoramic radiographs were identified as carious by TRIOS 5 (Fig. 4). Although the marginal difference between the two methods was not statistically significant (χ² = 3.760; p = 0.052), the calculated diagnostic performance measures suggested limited correspondence between TRIOS 5 and panoramic radiography (Tables 3, 4, 5). Using panoramic radiography as the comparator, the diagnostic performance indices of TRIOS 5 demonstrated low sensitivity, specificity, and overall discriminative ability (Table 6).

Fig. 3.

Panoramic radiographic and intraoral scanner model images of teeth 14 and 15 are presented. Approximal carious lesions were identified on the panoramic radiograph in both teeth (a). The lesions detected radiographically were also identified as carious lesions by the TRIOS 5 caries detection aid mode of scanner system (b). First molars were excluded from the evaluation due to the presence of existing restorations

Fig. 4.

Panoramic radiographic and intraoral scanner model image of teeth 35 and 45 are presented. Teeth are classified as non-carious on panoramic radiographs (a-b). 3D digital scan image of the mandible (c). Same teeth are identified as carious by TRIOS 5 caries detection aid mode (d)

Table 3.

Evaluation of TRIOS 5 for caries detection using panoramic radiography as the comparator method

Test: TRIOS 5	Comparator Method: Panoramic Radiography			χ2	p
	Caries Present	No caries	Total
Caries present	98	211	309	3,760	0,052
No caries	173	29	202
Total	271	240	511

χ2:McNemar chi-square test statistic

Table 4.

Mixed-effects logistic regression model for caries detection using patient as a random effect

X²	df	p
5.745	1.000	0.017

LogLikelihood=-696.17, AIC: 1398.40, BIC: 1413.13The estimated patient-level random intercept variance was 0.000 (SD = 0.000; ICC = 0.000). The residual variance was 3.290 (SD = 1.814), (Number of observations = 1,022; number of patients = 60)

Table 5.

Estimated marginal means (predicted probabilities) from the mixed-effects logistic regression model

			95% Confidence Intervals
Method	Mean	SE	Lower	Upper
Panoramic Radiography	0.470	0.02	0.427	0.51
TRIOS 5	0.395	0.022	0.354	0.438

Expected marginal means are expressed as probabilities

SE Standard error

Table 6.

Comparative diagnostic measures of TRIOS 5 using panoramic radiography as the comparator method

Comparative measures		95% CI
		Lower	Upper
Sensitivity	36,2%	30,4%	42,2%
Specificity	12,1%	8,2%	16,9%
Positive Predictive Value	31,7%	26,6%	37,2%
Negative Predictive Value	14,4%	9,8%	20,0%
Positive Likelihood Ratio	0,411	0,349	0,485
Negative Likelihood Ratio	5,283	3,712	7,518
Diagnostic Odds Ratio	0,078	0,049	0,123

CI Confidence Interval. The reported measures were calculated relative to panoramic radiography as the comparator method and should not be interpreted as estimates of absolute diagnostic accuracy

When the visual assessment of the 3D digital scan images was used as the comparator, TRIOS 5 demonstrated substantially improved diagnostic performance. The majority of carious lesions identified by visual assessment were also classified as carious by TRIOS 5, although a considerable proportion of visually sound teeth were classified as carious. The marginal difference between the two methods was statistically significant (χ² = 51.546; p < 0.001). Compared with panoramic radiography, higher sensitivity, specificity, negative predictive value, and diagnostic odds ratio were observed under this comparison.

Molar teeth

In molar teeth, the caries detection rate was 40.5% on panoramic radiographs and 72.2% with TRIOS 5. The level of agreement among the three assessment methods was low (Fleiss’ κ = −0.141; p < 0.001), whereas the highest pairwise agreement was observed between the visual assessment of the 3D digital scan images and TRIOS 5 (Gwet’s AC1 = 0.333; p < 0.001). Using panoramic radiography as the reference method, TRIOS 5 showed a sensitivity of 50.0%, specificity of 12.6%, positive and negative likelihood ratios of 0.572 and 3.957, respectively, and a diagnostic odds ratio of 0.145. When the visual assessment of the 3D digital scan images was used as the reference method, sensitivity increased to 81.3% and specificity to 40.0%, with positive and negative predictive values of 64.7% and 61.2%, respectively.

Premolar teeth

In premolar teeth, the caries detection rate was 69.2% on panoramic radiographic evaluation and 14.8% on visual assessment of the 3D digital scan images. Overall concordance among the three methods was poor (Fleiss’ κ = −0.270; p < 0.001), and pairwise comparisons also demonstrated negative agreement coefficients. Using panoramic radiography as the reference method, TRIOS 5 showed a sensitivity of 28.6%, specificity of 7.1%, a negative predictive value of 4.3%, a negative likelihood ratio of 10.000, and a diagnostic odds ratio of 0.031. When the visual assessment of the 3D digital scan images was used as the reference method, the sensitivity and specificity of TRIOS 5 were 70.4% and 55.5%, respectively, with positive and negative predictive values of 21.6% and 91.5%. Evaluation according to caries type demonstrated that panoramic radiography classified a greater proportion of lesions as approximal caries, whereas TRIOS 5 and the visual assessment of the 3D digital scan images more frequently classified lesions as occlusal caries (Table 7). Overall agreement among the three methods was low, and pairwise analysis demonstrated that the highest agreement was observed between TRIOS 5 and the visual assessment of the 3D digital scan images (Table 8).

Table 7.

Caries types according to the assessment methods

Caries Type	Panoramic radiography	TRIOS 5	3D Digital scan images
	n (%)	n (%)	n (%)
Occlusal	42 (15,5)	193 (62,5)	149 (71,3)
Aproximal	229 (84,5)	109 (35,3)	46 (22,0)
Buccal/palatinal	0 (0,0)	7 (2,3)	14 (6,7)
Total	271 (100,0)	309 (100,0)	209 (100,0)

Table 8.

Agreement levels between assessment methods according to caries type

Source	n	Observed Agreement (Po)		%95 CI	p
Panoramic Radiography - TRIOS 5	482	%12,7	-0,119G	-0,162 - -0,076	< 0,001
Panoramic Radiography − 3D digital scan images	401	%12,2	-0,127G	-0,172 - -0,077	< 0,001
3D digital scan images - TRIOS 5	356	%40,2	0,237G	0,156–0,321	< 0,001
Overall Agreement (Panoramic - TRIOS 5–3D images)	507	%6,3	-0,012F	-0,059 − 0,030	0,512

 Confidence intervals were obtained by a 1000-repetition patient-level clustered bootstrap percentile method. In pairwise analyses, observations classified as “no caries” by both methods were excluded; in the overall analysis, observations classified as “no caries” by all three methods were excluded

G Gwet’s AC1, F Fleiss’ kappa, CI Confidence interval

In molar teeth, evaluation according to caries type demonstrated that the agreement between visual assessment and TRIOS 5 was higher than that observed for the other method pairs (Gwet’s AC1 = 0.387; p < 0.001). In contrast, all comparisons involving panoramic radiography showed negative agreement coefficients. In premolar teeth, all pairwise comparisons demonstrated negative agreement coefficients, and the level of agreement remained considerably low.

Evaluation according to lesion depth revealed that the proportion of dentin-involved lesions was higher in panoramic radiographic assessment (84.5%) compared with the other methods. In the TRIOS 5 evaluation, the proportion of initial lesions was 46.9%, whereas this rate was determined as 64.6% in the visual assessment. The concordance between the three methods was low, with a Krippendorff’s alpha coefficient of − 0.136 (p < 0.001). Pairwise comparisons demonstrated markedly negative agreement coefficients between panoramic radiography and the other methods, whereas no statistically significant agreement was observed between visual assessment and TRIOS 5 (Gwet’s AC2 = − 0.089; p = 0.106).

Complementary mixed-effects analyses did not materially alter the overall interpretation of the findings, as patient-level clustering was negligible.

Discussion

In the present retrospective study, the fluorescence-based caries detection mode of the TRIOS 5 intraoral scanner was compared with panoramic radiography and ICDAS-based on-screen assessment for the detection of occlusal and approximal carious lesions. The fluorescence-based assessment showed higher detection rates for occlusal and enamel-limited lesions, whereas panoramic radiography identified a greater proportion of approximal and dentin-involved lesions. These findings suggest that differences between the evaluated methods are likely related to their different diagnostic principles rather than indicating superiority of one method over another.

One of the fundamental principles of minimally invasive dentistry is the early detection of initial caries lesions to allow preventive or micro-invasive treatment approaches [21]. However, conventional radiographic methods are known to have limitations, particularly in the detection of early enamel lesions. Systematic reviews conducted by Schwendicke et al. and Gimenez et al. reported that radiographic methods demonstrate high specificity but relatively low sensitivity for the detection of early lesions [5, 22]. This limitation is particularly evident on occlusal surfaces, since early demineralization within enamel generally does not become radiographically visible until progression into dentin occurs [2].

The higher detection rate of occlusal lesions observed with the fluorescence-based assessment in the present study was consistent with findings reported in previous studies evaluating the diagnostic potential of intraoral scanners. Michou et al. [17] demonstrated that a fluorescence-based intraoral scanner was capable of identifying occlusal carious lesions under clinical conditions. More recently, Agius et al. [18] reported encouraging diagnostic performance of a fluorescence-based intraoral scanner for caries detection on permanent tooth surfaces, although diagnostic performance varied according to lesion location and tooth surface. Similarly, Ntovas et al. [23] reported good agreement between occlusal caries assessment performed on TRIOS-derived three-dimensional digital models and clinical examination findings. Likewise, Schlenz et al. [24] showed that fluorescence-based caries detection modules integrated into intraoral scanners have the potential to support occlusal caries detection. The higher detection rate of occlusal and enamel-limited lesions observed in the present study is in line with these previous observations.

An important finding of the present study was the higher agreement observed between the fluorescence-based assessment and ICDAS-based on-screen assessment compared with panoramic radiography. This may be explained by the fact that both methods primarily evaluate surface morphology and optical alterations. While the ICDAS system is based on early changes in opacity, discoloration, and surface integrity, the fluorescence-based caries detection system evaluates fluorescence patterns associated with demineralization. Consequently, both methods primarily assess surface alterations. In contrast, panoramic radiography evaluates changes in mineralized tissues based on radiographic density, which generally become evident only after further lesion progression.

The present study showed that panoramic radiography identified a greater proportion of approximal and dentin-involved lesions than the fluorescence-based IOS assessment. This finding is consistent with the inherent characteristics of radiographic imaging, particularly for the evaluation of approximal surfaces, which cannot be directly visualized because of adjacent teeth [4]. In contrast, the fluorescence-based IOS assessment identified fewer approximal lesions in the present study. Kanar et al. [25] reported that a NIRI-based intraoral scanner achieved diagnostic performance comparable to bitewing radiography for the detection of approximal caries. However, their study differed substantially from the present investigation in both the imaging technology and study design, as it evaluated a near-infrared transillumination system using bitewing radiographs with histological validation, whereas the present study evaluated a fluorescence-based caries detection system using archived panoramic radiographs in a retrospective clinical setting. Accordingly, differences in both the radiographic comparator and the underlying optical technologies used for caries detection should be considered when interpreting the lower agreement observed in the present study.

Another notable finding of the present study was the better performance of the fluorescence-based assessment in molar teeth compared with premolars. This difference may be related to anatomical characteristics of molar teeth, including their broader occlusal surfaces and more complex pit-and-fissure morphology, which may facilitate the identification of occlusal lesions. In contrast, the relatively smaller occlusal surface area and different anatomical morphology of premolars may contribute to lower lesion detection by fluorescence-based assessment. Similar findings were reported by Kanar et al. [25], who observed higher diagnostic performance in molar teeth using a NIRI-based intraoral scanner. Although a different optical technology was evaluated in that study, the consistent trend across both investigations suggests that tooth morphology may influence the performance of optical caries detection systems.

The low agreement observed among the evaluated methods may reflect the different lesion characteristics assessed by each diagnostic approach. Panoramic radiography primarily evaluates mineralized tissue changes based on radiographic density, whereas the fluorescence-based IOS evaluates fluorescence changes associated with demineralization. ICDAS-based on-screen assessment, on the other hand, is primarily based on surface morphology and color changes. Consequently, the same lesion may be identified differently depending on the diagnostic principle of the method applied and the stage of lesion progression. A recent review by Chan et al. [26] similarly emphasized the potential role of fluorescence- and transillumination-based diagnostic systems in the detection of early carious lesions, while highlighting variability in image interpretation and the lack of standardized assessment protocols as important limitations.

The higher detection rate of occlusal and enamel-limited lesions observed with the fluorescence-based assessment should be interpreted with caution. Previous studies have reported that optical caries detection systems may be influenced by factors such as fissure pigmentation, surface discoloration, natural variations in enamel translucency, and optical artefacts, which may affect lesion interpretation [24, 26]. In addition, lesion visibility may vary according to tooth morphology, lesion depth, and lesion location, particularly on approximal surfaces where optical assessment is inherently more challenging. Therefore, fluorescence-based caries detection should be considered an adjunct to clinical examination and radiographic assessment rather than a standalone diagnostic approach.

One distinctive aspect of the present study was the application of ICDAS-based visual assessment to three-dimensional digital intraoral scan models. Compared with conventional clinical examination, these models can be magnified, viewed from multiple angles, and re-evaluated under standardized conditions. This digital approach may facilitate more consistent visual assessment while remaining consistent with the principles of minimally invasive dentistry, in which early detection of carious lesions plays a central role.

An important advantage of fluorescence-based intraoral caries detection systems is that they do not require ionizing radiation. This characteristic may be particularly beneficial when repeated examinations are clinically indicated or when minimizing radiation exposure is desirable. In addition, the integration of caries detection into routine digital intraoral scanning may provide additional diagnostic information without requiring an additional imaging procedure. Nevertheless, the findings of the present study indicate that fluorescence-based assessment should be considered a complementary approach to conventional clinical examination and radiographic evaluation.

The findings of the present study should be interpreted in light of several methodological limitations. First, because of the retrospective study design, an independent reference standard, such as histological or operative validation, was not available. Consequently, the present investigation primarily evaluated the agreement among different diagnostic methods rather than their absolute diagnostic accuracy. Furthermore, panoramic radiography served as the comparator method and should not be regarded as a definitive reference standard. Therefore, the calculated diagnostic performance measures should be interpreted together with the inter-method agreement analyses rather than in isolation.

The use of panoramic radiography was primarily determined by the retrospective archive-based design of the study. In university hospitals, panoramic radiographs are routinely acquired during patient care and are therefore readily available in institutional digital archives, whereas bitewing radiographs are obtained only when clinically indicated and were not consistently available for all patients included in the present dataset. Consequently, panoramic radiography represented the most appropriate comparator for this retrospective study, although differences between panoramic radiography and the fluorescence-based assessment should be interpreted in the context of their different diagnostic principles rather than as evidence of superiority of either method.

In addition, the single-center design and the inherent limitations of panoramic radiography, particularly for the detection of early and approximal carious lesions, may have influenced the findings. Nevertheless, the use of routinely archived panoramic radiographs and digital intraoral scan records reflects real-world clinical practice and provides valuable information regarding the agreement among these diagnostic approaches in a retrospective clinical setting. Future prospective studies incorporating independent reference standards, such as bitewing radiography, operative validation, or histological examination where appropriate, are warranted to further evaluate the diagnostic performance and clinical applicability of fluorescence-based caries detection integrated into intraoral scanners.

Conclusions

Within the limitations of this retrospective study, the fluorescence-based caries detection mode of the TRIOS 5 intraoral scanner and panoramic radiography provided complementary diagnostic information, reflecting their different diagnostic principles. In clinical practice, fluorescence-based findings should therefore be interpreted together with clinical examination and radiographic assessment. Further prospective studies using independent reference standards are warranted to confirm the diagnostic performance of fluorescence-based caries detection integrated into intraoral scanners.

Abbreviations

ICDAS

International Caries Detection and Assessment System

QLF

Quantitative Light—Induced Fluorescence

FOTI

Fiber Optic Transillumination

DIFOTI

Digital Imaging Fiber Optic Transillumination

OCT

Optical Coherence Tomography

NIRI

Near—Infrared Imaging

IOS

Intraoral Scanner

PACS

Picture Archiving and Communication System

AC1

Agreement Coefficient 1

AC2

Agreement Coefficient 2

CI

Confidence Interval

PPV

Positive Predictive Value

NPV

Negative Predictive Value

LR+

Positive Likelihood Ratio

LR−

Negative Likelihood Ratio

DOR

Diagnostic Odds Ratio

Authors’ contributions

F.O.T., B.E. and N.A.: Study design and drafting manuscript; F.O.T. and S.K.: Conceived the ideas and data analysis; F,O.T. and N.A.: Data collection; F.O.T., M.B. and S.K.: Performing statistical analysis; F.O.T. and N.A.: Concept and critical revision of the manuscript. All authors have read and approved the manuscript.

Funding

This research has not received any funding.

Data availability

The datasets used and/or analyzed during the current study are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

Ethical approval was obtained from the Institutial Ethics Committee of Health Sciences University Gülhane prior to data collection (Approval No: 2025 − 233). The study was conducted in accordance with the principles of the Declaration of Helsinki and all relevant guidelines and regulations. Written informed consent, including permission for the use of clinical data for scientific research purposes, was obtained from all participants.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.