Artificial Intelligence (AI) in Diagnosis, Artificial Intelligence Supported Image Reviewing
Conditions
Keywords
Artificial Intelligence, Caries Detection, Decision-making, Dental Imaging, Treatment Planning
Brief summary
This study aims to evaluate the influence of artificial intelligence (AI) on the decision-making process for intervention after caries lesion detection. Participants will be dentists working in the Netherlands randomly divided into two groups. Dentists will be divided into two groups and receive a set of bitewing radiographs, which first will be evaluated with or without AI support according to their group. Participants will examine caries lesions on the radiographs and formulate treatment plans accordingly. Then, after a wash-out period of one month, the same radiographs, but in the opposite condition of AI support and again formulate treatment suggestions according to the present caries lesions.
Detailed description
This crossover randomized controlled trial evaluates the effect of artificial intelligence (AI) decision support on dentists' treatment planning following caries detection bitewing radiographs. The study targets clinical decision-making processes by assessing how AI influences diagnostic interpretation and subsequent treatment suggestions. Dentists will be randomly assigned into two study arms. Each participant will evaluate a standardized set of digital bitewing radiographs under two conditions: once with AI assistance and once without, separated by a one-month wash-out period to minimize recall bias. The AI tool provides caries detection prompts based on radiographic analysis but does not suggest treatment. The crossover design enables within-subject comparison, controlling for individual diagnostic thresholds. The radiographs remain constant across both phases to isolate the influence of AI support. The study focuses on diagnostic performance and clinical decision outcomes, both with and without AI support. Treatment decisions are categorized into three predefined levels: no treatment, non-invasive treatment (e.g., fluoride application, polishing, sealing), and invasive intervention (i.e., restorative treatment). Diagnostic accuracy is measured against a reference standard and reported in terms of sensitivity and specificity. Caries detection will be classified using a modified International Caries Classification and Management System (ICCMS). This study design allows to quantify AI's impact on diagnostic performance, as well as on potential shifts in treatment approach. The study aims to contribute to evidence-based guidance on the integration of AI tools into clinical dental practice.
Interventions
DIAGNOSTIC_TESTArtificial intelligence in diagnosis
AI-based diagnostic programs have proved to enhance diagnostic performance, however research on its effects on treatment decisions is scarce. In contrast to other studies focusing on AI's accuracy or the resulting increase in dentists' accuracy, this study aims to investigate the differences in dentists' treatment recommendations when supported by AI versus when they are not during caries detection.
Sponsors
Prime Dental Alliance Eindhoven
Radboud University Medical Center
Study design
Allocation
RANDOMIZED
Intervention model
CROSSOVER
Primary purpose
DIAGNOSTIC
Masking
SINGLE (Subject)
Eligibility
Sex/Gender
ALL
Healthy volunteers
Yes
Inclusion criteria
1. Graduated, practising dentists. 2. At least three years of experience
Exclusion criteria
1. Retired dentists. 2. Specialized practitioners (e.g., orthodontists and oral surgeons) if their typical practice does not involve routine caries diagnostics and treatment planning.
Design outcomes
Primary
| Measure | Time frame | Description |
|---|---|---|
| Treatment decisions: Compare the treatment recommendations of dentists for caries lesions detected with and without AI support. | Each participant will be assessed over a period of up to 2 months (includes both evaluation phases and washout period) | The given options will be no treatment, non-invasive treatment (fluoride varnish, polishing, sealing), and restoration. Participants' answers will be compared to a reference standard. |
Secondary
| Measure | Time frame | Description |
|---|---|---|
| Diagnostic Accuracy in Caries Detection | Each participant will be assessed over a period of up to 2 months (includes both evaluation phases and washout period) | RA0: No radiolucency - No visible caries. RA1: Radiolucency confined to the outer half of enamel - Enamel caries. RA2: Radiolucency extending to the inner half of enamel but not reaching dentin - Moderate enamel caries. RA3: Radiolucency extending into the outer third of dentin - Dentin caries. RA4: Radiolucency extending into the middle of dentin - Advanced dentin caries. RA5: Radiolucency extending into the inner third of dentin - Severe dentin caries. Participants' answers will be compared to a reference standard. |
Countries
Netherlands
Outcome results
None listed