A dental scanner can produce a remarkably convincing three-dimensional picture, but can that picture actually detect decay around an old filling? This study finds a useful answer: sometimes—but especially as a complement, not a replacement.

Secondary caries lesions, also called caries lesions around restorations, are among the main reasons for restorative intervention, repair, or replacement, regardless of the restorative material. They are defined as lesions at the margins of existing restorations, or lesions associated with restorations or sealants.

These lesions develop around an existing dental restoration and are associated with marginal failures in the restoration, allowing bacteria to penetrate and initiate demineralization of the dental structure. Alongside tooth and restoration fractures, secondary caries lesions are the leading cause of restoration replacement, and if left undiagnosed and untreated, they may progress and lead to further damage of the tooth structure.

The diagnosis of secondary caries lesions presents a clinical challenge because an associated restoration is present. Currently, clinical examination combined with radiography is considered the conventional diagnostic method for secondary caries. The clinical exam evaluates restoration integrity on visible surfaces, while radiography allows observation of proximal lesions, including their presence, depth, and relationship to the pulp cavity.

Specific visual criteria, such as CARS, or Caries Associated with Restorations or Sealants, guide evaluation by focusing on restoration margins and signs of active demineralization. The study aimed to validate a three-dimensional intraoral scanner for detecting secondary caries lesions around restorations using image-based CARS-derived severity categories, compared with visual-tactile clinical examination.

The hypothesis was that IOS images would show diagnostic accuracy comparable to visual-tactile clinical examination, with performance varying according to lesion severity. The sample size calculation considered a twenty percent prevalence of secondary caries lesions, sensitivity of eighty-five percent, specificity of eighty-five percent, a precision of ten percentage points, and a ninety-five percent confidence interval.

Based on these data, a total of two hundred forty-five restorations were required for the study. The estimated prevalence of secondary caries was based on a review reporting restoration survival rates ranging from forty-three to ninety-nine percent, depending on study characteristics and observation period.

Because the study was conducted in a university clinical setting and included restorations with varying service times, most between five and ten years, an overall survival rate of approximately eighty percent was assumed. Three examiners performed the assessments: two performed the clinical assessment, and one performed the three-dimensional scanner image assessment.

Before evaluations, the examiners underwent training and calibration procedures, conducted in three phases. In the first phase, the scanner examiner was trained in equipment handling under the supervision of an experienced professional and practiced scanning techniques.

In the second phase, the image examiner received theoretical training using clinical photographs of restored teeth with various CARS scores for case discussion. In the third phase, all three examiners separately recorded diagnoses based on CARS criteria while viewing three-dimensional scanned images not included in the study.

Intraoral scans were performed using the TRIOS three three-dimensional scanner, supported by TRIOS software version one point eighteen point two point eleven and Dental Desktop version one point six point eight point one. All teeth were scanned under standard white light to produce three-dimensional models accurately reflecting natural tooth color, with the teeth kept moist throughout the procedure.

Images were saved in DCM and STL formats and exported for further analysis. Scans were considered adequate when the software captured sufficient detail of the examined teeth, strictly following the manufacturer’s guidelines.

Diagnostic accuracy was assessed at the restoration level, using clinical examination as the reference standard. The measures included sensitivity, specificity, area under the receiver operating characteristic curve, and positive and negative likelihood ratios.

Positive and negative predictive values were calculated alongside the other diagnostic measures. Because multiple restorations came from the same participant, ninety-five percent confidence intervals were estimated using participant-level cluster bootstrap resampling with two thousand replications.

The same procedure was applied consistently across all diagnostic thresholds. Because IOS classifications were binary, the area under the receiver operating characteristic curve was calculated as sensitivity plus specificity divided by two, corresponding to balanced accuracy.

The reported AUC values therefore represent binary diagnostic classification performance rather than a multi-threshold ROC analysis. Fifty-nine individuals were invited to participate, and eighteen did not meet the inclusion criteria because they had no direct restorations.

There were no refusals. A total of two hundred ninety-nine restorations were evaluated across forty-one patients: forty-six were in anterior teeth, and two hundred fifty-three were in posterior teeth. The patients were aged between eighteen and sixty and included thirty-four women and seven men.

The mean number of restorations per person was seven point two nine, ranging from two to seventeen. Figure one presents four representative intraoral scanner images of restored posterior teeth, labeled A through D. The images show different restoration appearances, including tooth-colored and darker restorative materials, with visible variation in occlusal surfaces and surrounding tooth structure.

These examples matter because the scanner examiner used three-dimensional images to assess caries around restorations while remaining blinded to the separate clinical evaluations. Table three presents the diagnostic properties of the intraoral scanner for caries lesions around restorations according to cumulative CARS disease severity thresholds.

Sensitivity was seventy-five percent at the initial lesion threshold, forty percent at the moderate lesion threshold, and fifty-seven point one percent at the severe lesion threshold. Specificity was eighty-four point one percent, ninety-seven point four percent, and ninety-nine point seven percent at the initial, moderate, and severe lesion thresholds, respectively.

The area under the ROC curve was zero point seven nine at the initial lesion threshold and zero point six nine at the moderate lesion threshold. Looking down the likelihood-ratio rows, the positive likelihood ratio rises from four point seventy-three at the initial threshold to fifteen point seven at moderate and one hundred sixty-six point eight at severe.

In the agreement row, kappa is zero point fifty-four, zero point forty-three, and zero point sixty-six across the same columns, while the confidence intervals show substantial uncertainty for several severe-threshold estimates. Although some studies have investigated three-dimensional intraoral scanners for diagnosing primary caries lesions, this was the first study, to the authors’ knowledge, to explore their clinical viability for detecting caries lesions around restorations using CARS visual criteria.

Overall, the IOS demonstrated satisfactory diagnostic capability in detecting secondary caries lesions across initial, moderate, and severe stages. Using CARS on IOS images represents an image-based adaptation rather than a direct reproduction of the clinical assessment.

This may have affected detection of lesions with subtle clinical features, particularly at the moderate lesion threshold, where tactile assessment or controlled wet and dry visualization may be needed for more precise classification. As a result, IOS sensitivity at the moderate lesion threshold may have been underestimated.

The findings support conventional color-textured three-dimensional intraoral scanner images as a complementary method for assessing caries lesions around restorations using image-based CARS-derived severity categories. Specificity remained high across lesion thresholds, whereas sensitivity varied according to lesion severity and was less precisely estimated at the severe threshold because of the small number of severe cases.

IOS may contribute to assessment and documentation of restored teeth, but its findings should be interpreted together with complete clinical examination and radiographic assessment when indicated. The scanner showed consistently high specificity, but sensitivity varied with lesion severity, especially for moderate lesions.

Its best role is complementary: documenting restorations while clinical examination and radiography remain part of the diagnostic picture.