Automatic identification of early bone loss patterns from radiographs invisible to human eyes for early periodontal disease diagnosis and prevention - Abstract: Periodontitis is the second most prevalent but preventable dental disease affecting over 64 million
Americans and responsible for tooth loss, functionality limitations, pain, and poor quality of life. Thus, early
diagnosis and preventive therapeutics are imperative in clinical practice to prevent disease initiation and
progression. However, by the time dentists can observe the first bone loss patterns in radiographs to diagnose
periodontitis, 30-50% deterioration (periodontal bone damage) has already occurred, which is not visible to
human eyes. Clinical decision support systems are designed to identify high-risk periodontitis patients for
prevention; however, they are not widely used in clinical practice because of the suboptimal prediction
performance and lack of diverse predictive features (early bone loss lesions) for prediction. Therefore, there is
an unmet need for a tool that can detect early bone loss patterns invisible to human eyes to alert dentists for
early diagnosis and preventive care. Dr. Patel has developed an artificial intelligence (AI) empowered prediction
model for periodontitis that utilizes more than 150 distinct variables (e.g., social determinants of health, medical
records, lab reports, CDC census data, financial data, etc.) for prediction, which aren't well understood in the
existing literature. However, this model lacks dental imaging data such as bone pattern, bone density, pixel
intensity, and other imaging predictive features, which have a high potential to improve prediction accuracy. The
early bone mineral changes in alveolar bone for early diagnosis have been studied in biological studies; however,
the transition of these findings at the chairside is limited. AI and computer vision can bridge this gap and help
identify early bone loss patterns from radiographs invisible to human eyes. Therefore, the objective of this project
is to develop three automated computer vision algorithms: 1) to improve the extraction of diagnostically
meaningful information from periapical radiographs, 2) to determine the extent of bone loss information from
radiographs, and 3) build a prediction model to identify early bone loss patterns from radiographs before disease
initiation and progression. Enhanced and consistent radiographs will improve diagnostic accuracy & reduce
radiographic exposure, automatic bone loss measurement will reduce diagnostic discrepancies, and early bone
loss detection will identify high-risk patients to take preventive approaches. The candidate, Dr. Patel's goal is to
become an independent PI in dental informatics and develop cutting-edge technologies to generate practice-
based evidence (using data-driven methods) to improve patient care and outcomes. A funded K08 proposal will
allow Dr. Patel to develop the skills necessary to complete the proposed research (training in computer vision &
radiology) and become an independent research scientist (training in didactic mentoring, lecturing, &
grantsmanship). Dr. Patel has formed a team of five mentors with expertise in clinical dentistry, computer vision,
radiology, and periodontology to provide high-quality, diverse scientific, collegial support and state-of-the-art
facilities to ensure the successful completion of this proposed career development goals and research program.
