GenAI-driven machine learning for periodontitis prediction using systemic indicators: AUC 0.48–0.57, SVM sensitivity 89% for screening

Retrospective study: in 416 dental patients, a GenAI pipeline trained 6 ML models on systemic indicators for periodontitis prediction. AUCs were modest (0.48–0.57); SVM had 89% sensitivity, suggesting screening utility where radiography unavailable, but diagnostic accuracy limit…

• Retrospective single-center study of 416 patients, 80/20 split with fivefold cross-validation; models used only systemic/demographic variables, no dental radiography.
• Highest sensitivity for screening: SVM 89%; logistic regression most balanced with 72% accuracy and 74% F1-score; overall AUCs 0.48–0.57.
• Limited by single-center design, poor discriminative ability (AUC near chance), and lack of imaging; not a diagnostic tool but a potential triage flag.

RadPigeon summaries are original and for information only. They are not clinical advice.