Predict: advancing periodontitis care with artificial intelligence (AI)-driven diagnostics and clinical decision support project summary periodontitis, a prevalent inflammatory disease affecting 42%…

The National Institute of Dental and Craniofacial Research awarded the University of Texas Health Science Center at Houston $1,187,388 on May 15, 2026, under the Oral Diseases and Disorders Research program (CFDA 93.121) to develop, validate, and clinically evaluate PREDICT, an artificial intelligence-driven clinical decision support system for periodontitis diagnosis and risk stratification.

Periodontitis affects 42 percent of U.S. adults and carries estimated misclassification rates of 32 percent, which delay appropriate treatment and worsen patient outcomes. Current diagnostic approaches rely on manual interpretation prone to inter-examiner variability. PREDICT integrates multi-modal dental imaging (2D and 3D) with machine learning algorithms trained on longitudinal electronic health record data to improve diagnostic accuracy and predict disease development and progression. The recipient will validate and adapt these AI models using retrospective and prospective data from Willamette Dental, a large multi-office dental practice network, and conduct a phased pilot study of the deployed clinical decision support tool in real-world dental practice.

Work is performed in Houston, Texas, with a performance period through February 28, 2031. The Skourtes Institute received a sub-award (R01DE035894-SA0005496) on July 12, 2026, supporting subordinate components of the project.

Predict: advancing periodontitis care with artificial intelligence (AI)-driven diagnostics and clinical decision support project summary periodontitis, a prevalent inflammatory disease affecting 42% of US adults, poses a significant public health challenge due to its complex diagnosis and high rates of misclassification (estimated at 32% in our recent findings). Such errors impede timely and appropriate treatment, leading to poorer patient prognoses. Current diagnostic paradigms rely heavily on manual interpretation and are prone to inter-examiner variability. This project confronts this critical issue by proposing the development, validation, and clinical evaluation of an artificial intelligence (AI)-driven system, predict (periodontitis recognition and estimation using diagnostic intelligence & computational techniques). Our central objective is to improve the accuracy of periodontitis diagnosis and risk stratification, culminating in a clinical decision support (CDS) tool designed for real-world dental practice. We will achieve this through three synergistic aims. Aim 1 will focus on creating AI models that integrate multi-modal dental imaging (2D and 3D) to deliver precise, image-based diagnostic assessments. These models will be validated and adapted using extensive retrospective data from willamette dental (WD), a large, multi- office dental practice network. Aim 2 will develop advanced machine learning algorithms using longitudinal electronic health record (EHR) data to accurately predict periodontitis development and progression, accounting for individual patient clinical trajectories. These progression models will also undergo validation and adaptation with WD's EHR data. Finally, aim 3 will leverage the validated and refined models from aims 1 and 2 to develop, deploy, and conduct a prospective, phased pilot study of the deployed AI-driven CDS tool. This tool will be informed by our team's prior work on the precision periodontal health care chart (PPHCC) at uthealth and adapted to WD's clinical environment, which utilizes similar foundational concepts for risk assessment and treatment planning. This evaluation will critically assess the predict-CDS's impact on diagnostic accuracy, progression prediction, clinician decision-making, and integration into existing clinical workflows. This research is innovative in its comprehensive approach: from foundational AI model development using diverse data sources to piloting of an integrated CDS built upon established principles and adapted for a large clinical network. We expect the predict-CDS to provide a transformative decision support tool for oral healthcare providers, significantly enhancing the quality, accuracy, and efficacy of periodontal diagnosis and treatment planning. Ultimately, this project aims to reduce diagnostic errors, personalize patient care, and improve oral health outcomes on a broad scale.