Unveiling AI-Augmented Periodontal Decision-Making: A Comparative Clinical Evaluation of Diagnocat-Based Radiographic Deep Learning and ChatGPT-Driven Treatment Planning Versus Clinical Expert Judgement

Abstract

Objective: To evaluate the agreement between periodontal treatment plans generated using a hybrid artificial intelligence (AI) workflow integrating Diagnocat and ChatGPT and clinician-derived consensus treatment plans, and to compare the performance of cone-beam computed tomography (CBCT)-based and orthopantomogram (OPG)-based AI-assisted treatment-planning workflows in patients with Stage II–IV periodontitis. Methods: This retrospective study included radiographic records of 60 patients diagnosed with Stage II, III, or IV periodontitis. For each patient, CBCT and OPG radiographs were independently analyzed using Diagnocat, an AI-based radiographic interpretation platform. Structured diagnostic reports generated by Diagnocat were converted into standardized clinical summaries and submitted to ChatGPT for generation of evidence-based periodontal treatment plans. Three experienced periodontists independently evaluated the same cases and established consensus treatment plans, which served as the reference standard. Agreement, sensitivity, specificity, accuracy, Cohen's kappa coefficient (κ), receiver operating characteristic (ROC) curve analysis, and area under the curve (AUC) were calculated using IBM SPSS Statistics Version 27.0. Results: The CBCT-based AI workflow demonstrated higher agreement with clinician-derived consensus treatment plans than the OPG-based workflow (76.7% vs. 65.0%). Agreement between AI-generated and clinician-derived treatment plans was higher for the CBCT-based workflow (κ = 0.52) than for the OPG-based workflow (κ = 0.38). The CBCT-based workflow achieved a sensitivity of 82%, specificity of 74%, and overall accuracy of 76.7%. Receiver operating characteristic analysis demonstrated an area under the curve (AUC) of 0.78 (95% confidence interval: 0.68–0.88), indicating acceptable treatment-planning performance. Conclusion: The integration of Diagnocat and ChatGPT demonstrated promising performance as an AI-assisted periodontal treatment-planning workflow. CBCT-based analysis showed superior agreement and treatment planning performance compared with OPG-based analysis, highlighting the value of three-dimensional imaging in AI-assisted clinical decision support. Although the hybrid AI workflow demonstrated encouraging results, clinician oversight remains essential for periodontal treatment planning.