Artificial Intelligence-Assisted Analysis Of Plaque-Induced Peri-Implant Diseases

The study aims to develop an AI-deep learning model for effcient intraoral plaque monitoring in dental implant patients and investigate plaque characteristics' correlation with clinical parameters to establish plaque as a predictive biomarker for peri-implant diseases, enabling early prevention and clinical decision support. Data from implant patients (Jan 2022-Jan 2024) at our hospital were analyzed. Standardized photos after plaque-disclosing agent application and intraoral 3D scans documented plaque distribution. Panoramic radiographs tracked post-implant bone loss. A trained dentist evaluated peri-implant health using gingival/bleeding indices and probing depth. AI development involves two phases: First, YOLO v8-based CNN for initial plaque detection from intraoral photos; second, octree-algorithm-enhanced CNN refinement. The system integrates radiographic bone loss data, clinical indices, and 3D plaque mapping to analyze plaque's role in peri-implant diseases. The study included 82 patients, yielding 246 intraoral photos, 82 3D scans, and 164 panoramic films. The AI model achieved 91.11% Dice/84.46% IoU scores for plaque detection on implants and natural teeth, outperforming existing models by 7%/11% with >50% faster processing. Results showed minimal plaque around implants triggered peri-implant inflammation more readily than natural teeth. Plaque accumulation at implant crown’s necks emerged as a critical risk factor for peri-implant diseases. Our AI algorithm detects plaque biofilm on teeth/implants, distinguishes their surfaces, links plaque to peri-implant diseases, and provides targeted hygiene guidance with early warnings.