Edentulous space classification and planning of clasp-retained removable dentures using generative artificial intelligence.

OBJECTIVE: This study examined whether generative artificial intelligence is able to identify and classify gaps in partially dentate jaws according to the Kennedy system and subsequently plan clasp-retained removable dentures. METHOD AND MATERIALS: Data from 100 partially dentate jaws were extracted from the patient records management system at the Department of Prosthetic Dentistry of the University Medical Center Hamburg-Eppendorf, Germany, and anonymized. The generative artificial intelligence tools that were included in the study comprised OpenAI's customizable 'GPTs' (GPT-4o model) and 'o1-preview.' After prompt engineering and training the GPTs with subject-specific information, three prompts were sent to the models stating the missing teeth in one arch and requesting a gap analysis, Kennedy classification, and planning of clasp-retained removable dentures. The prosthetic planning was also handled by a dentist. A statistical analysis was performed using Cohen's kappa, McNemar's test, chi-square test, and standardized residuals (significance level: .05). RESULTS: o1-preview correctly identified the gaps in significantly more cases than GPTs (83% vs 14%, P .001) and made fewer and different types of errors. o1-preview correctly classified more cases according to their Kennedy class (90%) than GPTs (40%). The degree of agreement with the dentist's treatment plan was 1% for o1-preview and 6% for the GPTs. CONCLUSION: While ChatGPT is not yet capable of reliably planning treatment, these results demonstrate its potential for use in dental diagnostics. For effective integration into clinical workflows, automated workflows through software solutions in dental practices, as well as the option of enhanced prompt insertion, are essential.