AI-Guided Cbct Diagnosis of Oral Lesions for Periodontal and Prosthodontic Planning

Abstract

Cone-beam computed tomography (CBCT) provides 3D assessment of oral lesions and periodontal/peri-implant defects, but its interpretation is time-consuming and observer dependent. Thus the aim of our study was to evaluate the diagnostic accuracy and clinical utility of an AI-guided CBCT analysis system for detecting oral lesions and periodontal/peri-implant defects, and to assess the AI-derived findings influence periodontal and prosthodontic treatment planning among 50 consecutive adult patients (412 tooth and implant sites) with an existing clinical indication for CBCT for periodontal, prosthodontic or combined assessment with the help of anonymised DICOM volumes using CNN & generate probability scores and segmentation overlays for horizontal and vertical bone loss, furcation involvement, periapical lesions, peri-implant marginal bone loss, ridge deficiency and other osseous lesions with 2 calibrated examiners. We have found that, diagnostic accuracy was 86.0% (95% CI 73.8–93.0) at patient level and 89.1% (85.7–91.7) at site level, with sensitivity 87.1% and 86.9%, specificity 84.2% and 91.1%, and substantial agreement (κ = 0.706 and 0.781); McNemar was non-significant. AI assistance raised clinician sensitivity from 75.9% to 86.9% (p < 0.001) and reduced reporting time by 37.1%. Treatment plans changed in 14 patients (28.0%, p < 0.001) respectively. Thus, we have come to conclude that, AI-guided CBCT is an accurate, clinician-supervised adjunct that meaningfully influences periodontal and prosthodontic planning.