Accuracy to detect dental caries on bitewings Gaze pattern and VAS-scale of the participants
Conditions
Interventions
Group 1: AI based caries detection on bitewings. Dentists will randomly receive radiographs from arm 1 and arm 2
this randomness will mean that a dentists may assess 1 radiograph using AI, then one without, then one with etc, at random.
Group 2: conventional caries detection on bitewings
Sponsors
Charité Berlin
Eligibility
Sex/Gender
All
Age
25 Years to No maximum
Inclusion criteria
Inclusion criteria: Dentists with more than 2 years of clinical experience (e.g. finished postgraduate education according to German insurance law, allowed to work self-employed and independently)
Exclusion criteria
Exclusion criteria: Dentists - no longer clinically active - no regular experience with caries detection (e.g. orthodontists / surgeons)
Design outcomes
Primary
| Measure | Time frame |
|---|---|
| Accuracy of AI-based X-ray diagnostics compared to conventional X-ray diagnostics. The participants findings will be compared against the gold standard that was defined by five dentists prior to data collection. An Excel sheet will be used for this | — |
Secondary
| Measure | Time frame |
|---|---|
| 1) Sensitivity, specificity, positive and negative predictive value, F1 score and Matthew's correlation coefficient, against the gold standard. 2) Intensity of non- and micro-invasive care, measured as assigned number of local applications of fluoride varnish, sealing, caries infiltration etc per viewed tooth. 3) Intensity of invasive care, measured as assigned restorations or further invasive care per viewed tooth. 4) Efficiency, measured as time from starting the view of the image to saving the final report. 5) Cost-effectiveness. A model-based evaluation of initial and follow-up costs will be performed under a mixed public private payers’ perspective in German healthcare. Costs will be estimated using fee items of the German public and private insurance. As effectiveness measure, the number of avoided invasive treatments and the time of tooth retention will be used. Discounting will be applied at 3% per annum. Opportunity costs will not be considered. A Markov model, constructed in previous studies, will be employed and analyzed using Monte Carlo microsimulations. Incremental cost-effectiveness ratios will be estimated. Univariate and probabilistic uncertainty analyses will be applied, and the net benefit approach used to construct cost-effectiveness-acceptability curves. 6) The diagnostic confidence will be recorded per image on a VAS scale from 1-10. 7) Process evaluation: For the process evaluation, a study protocol is prepared in which the questionnaires, interview guidelines, SOPs and evaluation strategies are specified. The perspectives of dentists as users will be explored. The aim of process evaluation is to make the effects and side effects of the intervention understandable. The methodological basis is provided by the guidelines of Moore et al (2015) and Grant et al (2013). It is necessary to develop a theory-based model of change (Theory of Change) in order to depict contextual factors (e.g. framework conditions, organizational structures) | — |
Countries
Germany
Contacts
Public ContactFalk Schwendicke
Charité Berlin
Outcome results
None listed