UTHealth Houston researchers awarded $2.8M NIH grant to develop AI system for periodontal disease diagnosis and prediction

UTHealth Houston researchers from the School of Dentistry and the McWilliams School of Biomedical Informatics are developing an AI tool to improve periodontal disease diagnosis and predict patient-specific risks. Photo by Kyle Rogers.

UTHealth Houston researchers from the School of Dentistry and the McWilliams School of Biomedical Informatics have been awarded a $2.84 million R01 grant from the National Institutes of Health to develop an artificial intelligence system designed to improve the diagnosis of periodontal disease and predict patient-specific disease progression risk using dental imaging and longitudinal electronic health record data.

The project, titled “PREDICT: Advancing Periodontitis Care with Artificial Intelligence-Driven Diagnostics and Clinical Decision Support,” brings together principal investigators, including Chun-Teh Lee, DDS, MS, DMSc, professor in the Department of Periodontics and Dental Hygiene and director of pre- and post-doctoral research at the School of Dentistry; contact principal investigator Muhammad Walji, PhD, professor and chair of the Department of Clinical and Health Informatics at McWilliams School of Biomedical Informatics, who also holds a dual appointment with the School of Dentistry; Xiaoqian Jiang, PhD, professor, associate vice president for medical AI at UTHealth Houston, and chair of the Department of Health Data Science and Artificial Intelligence; and Shayan Shams, PhD, assistant professor in the Department of Health Data Science and Artificial Intelligence. There are also two co-investigators, Oluwabunmi Bunmi Tokede, BDS, MPH, DMSc, at the School of Dentistry, and Lishan Yu, PhD, at McWilliams School of Biomedical Informatics.

Periodontitis affects an estimated 42% of U.S. adults, according to the Centers for Disease Control and Prevention’s National Health and Nutrition Examination Survey, and remains challenging to diagnose consistently due to its complexity and reliance on manual interpretation of radiographs, periodontal charting, and clinical findings.

Prior research by the team suggests that misclassification rates may be as high as 32%, contributing to delayed or less tailored treatment planning and variable patient outcomes.

The collaboration began in 2019, when Jiang, Shams, and Lee started developing deep learning models to interpret dental radiographs for periodontitis diagnosis. In 2021, the team expanded with Walji to integrate imaging and clinical data, advancing toward an artificial intelligence system capable of both diagnosis and prediction of disease progression. Over several years, the group has published multiple journal articles and conference proceedings supporting this work.

The PREDICT project will develop and validate artificial intelligence models that integrate multi-modal dental imaging, including two-dimensional and three-dimensional radiographs, with longitudinal electronic health record data to support more accurate diagnosis, individualized risk stratification, and earlier treatment planning.

The goal is to shift periodontal care toward a more proactive approach that identifies risk earlier in the disease process.

“Improved prediction would help clinicians move from a largely reactive model to a more proactive, risk-stratified approach,” Lee said. “If we can identify patients who are more likely to progress, clinicians can tailor monitoring, preventive care, and treatment intensity earlier, before more severe tissue and bone loss occurs.”

The resulting models will be used to develop a clinical decision support tool for dentists and periodontists. The system will be evaluated in partnership with Willamette Dental in a phased pilot study within a real-world clinical environment.

“Current practice relies heavily on clinicians manually interpreting radiographs and synthesizing clinical findings, which can vary across examiners and clinical settings,” Walji said. “PREDICT is designed as a clinical decision support system that combines imaging with electronic health record-based clinical trajectories to generate patient-specific diagnostic and progression-risk estimates.”

Rather than relying on static image interpretation alone, the approach extends the analysis across time.

“This project moves beyond isolated image analysis toward a more comprehensive understanding of periodontal disease over time,” Shams said. “That broader approach is what gives AI the potential to meaningfully support clinical decision-making.”

That time-based perspective is central to understanding how periodontal disease develops in real patients.

“The longitudinal component is especially important because disease progression is not just a single image interpretation problem; it is a pattern over time,” Jiang said. “By integrating imaging with longitudinal clinical data, we aim to build models that better reflect how periodontal disease develops and progresses in real patients.”

Researchers will evaluate the system’s ability to improve diagnostic accuracy, predict disease progression, and support clinical decision-making while integrating into existing dental workflows. Rather than acting as a stand-alone image reader, the tool is designed to combine imaging and clinical history into patient-specific risk estimates that clinicians can use alongside examination findings.

The ultimate aim is to reduce diagnostic variability, enhance treatment planning, and improve consistency in periodontal care through scalable artificial intelligence tools.

The project represents an important milestone in the use of artificial intelligence in dental medicine and marks the first major NIH-funded AI grant for UTHealth Houston School of Dentistry. The R01 support allows the team to move beyond retrospective model development to validation, deployment, and prospective pilot testing in real clinical workflows.