Trial · NCT07907900
Diagnostic Accuracy of Large Language Models for Peri-Implant Marginal Bone Loss on Panoramic Radiographs
NCT07907900, Not yet recruiting, Ankara University
3 sourcestrials-2026-11-12-0.5.0
Cite this record
ConductScience Library. Diagnostic Accuracy of Large Language Models for Peri-Implant Marginal Bone Loss on Panoramic Radiographs. Trials, record NCT07907900. Release trials-2026-11-12-0.5.0, 12 Nov 2026. https://conductscience.org/library/trials/NCT07907900
On this page
Trial
As registered on ClinicalTrials.gov, in the graph's capture of the registry.
- Title
- Diagnostic Accuracy of Large Language Models for Peri-Implant Marginal Bone Loss on Panoramic Radiographs
- NCT number
- NCT07907900
- Status
- Not yet recruiting
- Enrollment
- 784 participants, estimated
- Lead sponsor
- Ankara University
- Conditions
- Peri-implantitis, Alveolar Bone Loss, Dental Implant
- Interventions
- Multimodal large language model evaluation (ChatGPT, Gemini, Claude, DeepSeek, Grok)
- First posted
- 2026-10-30
- Last updated
- 2026-10-30
Source ClinicalTrials.gov
Registered outcomes5
Each outcome the trial registered, primary first, in ClinicalTrials.gov's words.
| List | Outcome | Time frame | Description |
|---|---|---|---|
| Primary | Sensitivity and specificity of each large language model for detecting peri-implant marginal bone loss | Single time point, at completion of model evaluation (approximately 8 weeks after study start); retrospective images, no follow-up | For each of the five models (ChatGPT, Gemini, Claude, DeepSeek, Grok): sensitivity and specificity for the presence of marginal bone loss (Froum and Rosen Score 1-3 vs Score 0) on anonymized panoramic ROI images, calculated against the consensus reference standard of three blinded clinicians, with 95% confidence intervals. |
| Secondary | Positive predictive value, negative predictive value and overall accuracy of each model | Single time point, at completion of model evaluation (approximately 8 weeks after study start) | PPV, NPV and overall accuracy of each of the five models for detecting marginal bone loss (present/absent), calculated against the reference standard with 95% confidence intervals. |
| Secondary | Agreement between model severity grading and the reference standard | Single time point, at completion of model evaluation (approximately 8 weeks after study start) | Agreement of the 4-level Froum and Rosen scores (0-3) assigned by each model with the consensus reference score, assessed with weighted kappa; disagreement between model and reference classification tested with McNemar's test. |
| Secondary | Area under the ROC curve of each model | Single time point, at completion of model evaluation (approximately 8 weeks after study start) | Discriminative performance of each model for marginal bone loss (present/absent), expressed as area under the receiver operating characteristic curve with 95% confidence interval. |
| Secondary | Inter-rater reliability of the three blinded clinicians | Single time point, at completion of clinician scoring (approximately 6 weeks after study start) | Fleiss' kappa for the independent Froum and Rosen scores (0-3) assigned by the three clinicians before consensus. |
Source ClinicalTrials.gov
Outcomes0
Each outcome the trial posted results for, in ClinicalTrials.gov's words.
None in this release.
Results by group0
Each measurement the trial posted, group by group, in the words and numbers it posted.
None in this release.
Participant flow0
How many started, completed or left each group, period by period.
None in this release.
Adverse events0
Each adverse event the trial posted, group by group: the participants affected out of those at risk.
None in this release.
Adverse event totals0
The totals the trial posted for each group.
None in this release.
Papers2
The papers the trial lists on ClinicalTrials.gov.
Grants0
The grants NIH RePORTER links to the trial, newest funded first: NIH's, and some of AHRQ's, CDC's, FDA's and VA's. Each shows its title, principal investigators, organization and funder in its newest year.
None in this release.
Trial comparisons0
Comparisons of the trial's two groups, calculated by ConductScience from the counts it posted.
None in this release.
Tests0
The ConductCognition tests this trial's registered outcomes name, as ScienceGraph matched them in trial-tests-2026-11-12-0.2.1.
None in this release.
Data dates
The day each source was copied for this release.
Sources 3
- ClinicalTrials.gov's posted results, through CTTI's AACTFreely available. Acknowledgement requested
- ClinicalTrials.govNo license stated
- NIH RePORTER's clinical studies and project filesPublic domain
Each source's terms
- ClinicalTrials.gov's posted results, through CTTI's AACTClinicalTrials.gov data via Aggregate Analysis of ClinicalTrials.gov (AACT), CTTI. CTTI makes its materials freely available and requests acknowledgement. https://aact.ctti-clinicaltrials.org/documentation/299
- ClinicalTrials.govnone stated
- NIH RePORTER's clinical studies and project filesPublic domain (NIH RePORTER)
ClinicalTrials.gov's data carry an international copyright outside the United States: https://clinicaltrials.gov/about-site/terms-conditions.
For software
Read this trial from software.
The record and each of its lists are JSON, twenty rows a page. No key.
GET https://conductscience.org/api/trials/v1/NCT07907900
GET https://conductscience.org/api/trials/v1/NCT07907900/registered