Artificial Intelligence for Radiographic Diagnosis of Peri-Implantitis: A Comprehensive Review on Detection,...

Area of Science:

• Dentistry
• Medical Imaging
• Artificial Intelligence

Background:

• Peri-implantitis poses a significant challenge in implant dentistry, with current radiographic diagnosis methods being operator-dependent and often detecting bone loss late.
• Artificial intelligence (AI) offers potential for enhanced detection, quantification, and prognosis of peri-implant bone conditions.

Purpose of the Study:

• To review and synthesize existing evidence on AI-based radiographic approaches for diagnosing peri-implantitis.
• To evaluate AI's role in detecting, measuring marginal bone loss, and stratifying risk associated with peri-implantitis.

Main Methods:

• A systematic literature search was conducted on PubMed and Scopus for studies from 2013-2025 using AI for peri-implantitis radiographic assessment.
• Ten studies were included, focusing on AI models applied to periapical, panoramic, or combined imaging for bone loss assessment.
• Data on imaging modality, AI architecture, task, dataset, performance metrics, and validation were qualitatively synthesized.

Main Results:

• AI models demonstrated encouraging performance in implant localization, marginal bone loss detection, keypoint identification, and severity classification.
• Periapical/intraoral radiographs were predominantly used, with common AI architectures including YOLO, Faster R-CNN, Mask R-CNN, U-Net, ResNet, and AlexNet.
• Only one study explored outcome prediction, and all included studies were retrospective with internal validation.

Conclusions:

• AI shows potential as an adjunctive tool for radiographic detection and quantification of peri-implant bone loss.
• Current evidence is limited by retrospective study designs, varied reference standards, and a lack of external validation and clinical data integration.
• Future research should focus on prospective, multicenter validation, longitudinal imaging, and multimodal AI models for robust clinical application.