Caries Diagnosis on Radiographs – A Retrospective Analysis
Conditions
Interventions
Group 1: Assessment of anonymized bitewing radiographs (T0/T1) using an AI-assisted analysis (Nostic) on approximal surfaces (mesial/distal) according to D0–D4
analysis performed blinded following randomized renaming/permutation of the image files.
Group 2: Independent assessment of the same anonymised bitewing radiographs by human examiners—two per study ce
Sponsors
Universitätsklinikum Marburg
Eligibility
Sex/Gender
All
Inclusion criteria
Inclusion criteria: Patients are included if two consecutive bitewing radiographs (T0 and T1) are available. At T0, at least two non-restored approximal surfaces (mesial or distal) must be present.
Exclusion criteria
Exclusion criteria: No exclusion criteria are defined.
Design outcomes
Primary
| Measure | Time frame |
|---|---|
| Re-detection rate of carious lesions and caries stages by Nostic between T0 and T1 Proportion (%) of carious lesions and caries stages identified by Nostic at T0 that are re-detected at T1 (same tooth, same region). Measurement method: automated lesion detection by Nostic; comparison of findings at the tooth level and lesion level. Time frame: T0 = initial radiograph, T1 = follow-up radiograph of the same patient. Re-detection rate of carious lesions by human examiners between T0 and T1 Proportion (%) of carious lesions identified at T0 that are re-detected at T1 (same region, same tooth). Comparison of Nostic vs. human examiners Agreement rate (%) for lesion re-detection between Nostic and the findings of human examiners. Measurement method: blinded assessment of the radiographs by dentists; calculation of sensitivity, specificity, and agreement (?). | — |
Secondary
| Measure | Time frame |
|---|---|
| Effect of caries stage Probability (%) of re-detection at T1 depending on the caries stage at T0 (E1, E2, D1, D2, etc.). Measurement method: stratified analysis by caries stage. Effect of overlap Difference in re-detection rate (%) between lesions located in overlapped vs. non-overlapped approximal areas. Measurement method: classification of radiographs by degree of overlap; comparison of re-detection rates. Effect of confidence score Association between the confidence score at T0 and re-detection of the lesion at T1. Measurement method: logistic regression. | — |
Countries
Germany
Contacts
Public ContactCarolina Ganß
UniversitätsZahnMedizin Marburg
Outcome results
None listed