Farina Roberto, Simonelli Anna, Trombelli Leonardo, Ettmayer Johanna B, Schmid Jan L, Ramseier Christoph A
Research Centre for the Study of Periodontal and Peri-Implant Diseases, University of Ferrara, Ferrara, Italy.
Operative Unit of Dentistry, Azienda Unità Sanitaria Locale (AUSL), Ferrara, Italy.
J Clin Periodontol. 2025 Aug;52 Suppl 29:211-245. doi: 10.1111/jcpe.14156. Epub 2025 Mar 19.
To comprehensively review digital technologies (including artificial intelligence, AI) for periodontal screening, diagnosis and prognosis in the dental setting, focusing on accuracy metrics.
Two separate literature searches were conducted for periodontal screening and diagnosis (part I, scoping review) and prognosis (part II, systematic approach). PubMed, Scopus and Embase databases were searched.
In part I, 40 studies evaluated AI and advanced imaging on different substrata. The combination of AI with 2D radiographs was the most frequently investigated and demonstrated a high level of periodontitis detection and stage definition. In part II, eight studies, identified as having a high risk of bias, tested supervised machine learning models using 6-74 predictors. The models demonstrated variable predictive accuracy, often outperforming traditional risk assessment tools and classical statistical models in the few studies evaluating such comparisons.
AI and advanced imaging techniques are promising for periodontal screening, diagnosis and prognosis in the dental setting, although the evidence remains inconsistent and inconclusive. In addition, AI-driven analysis of 2D radiographs (for diagnosis and staging of periodontitis), neural networks and the aggregation of multiple algorithms (for predicting tooth-related outcomes) appear to be the most promising approaches entering clinical application.
全面综述牙科环境中用于牙周筛查、诊断和预后的数字技术（包括人工智能，AI），重点关注准确性指标。
针对牙周筛查与诊断（第一部分，范围综述）和预后（第二部分，系统方法）分别进行了两次文献检索。检索了PubMed、Scopus和Embase数据库。
在第一部分中，40项研究评估了人工智能和先进成像技术在不同层面的应用。人工智能与二维X线片的结合是研究最频繁的，并且显示出高水平的牙周炎检测和分期定义。在第二部分中，八项被确定为存在高偏倚风险的研究，使用6至74个预测因子测试了监督式机器学习模型。在少数评估此类比较的研究中，这些模型显示出不同的预测准确性，通常优于传统风险评估工具和经典统计模型。
人工智能和先进成像技术在牙科环境中的牙周筛查、诊断和预后方面很有前景，尽管证据仍然不一致且尚无定论。此外，人工智能驱动的二维X线片分析（用于牙周炎的诊断和分期）、神经网络以及多种算法的汇总（用于预测与牙齿相关的结果）似乎是最有前景的进入临床应用的方法。
