Related Experiment Video
Updated: Jan 18, 2026
06:14
Induction of Periodontitis via a Combination of Ligature and Lipopolysaccharide Injection in a Rat Model
Published on: February 17, 2023
5.1K
Deep Learning-Based Detection of Periodontal Infrabony and Furcation Defects on Periapical Radiographs: A Feasibility
Nicola Alberto Valente1, Lorenzo Maria Americo2, Fabrizio Ciancetta3
1Division of Periodontics, School of Dental Medicine, Department of Surgical Sciences, Faculty of Medicine, University of Cagliari, Cagliari, Italy; College of Dentistry, American University of Iraq Baghdad (AUIB), Baghdad, Iraq.
International Dental Journal
|January 16, 2026
Summary
This study shows artificial intelligence (AI) can detect periodontal defects on X-rays, but accuracy is limited. AI models offer potential for consistent diagnosis and surgical planning in periodontology.
Area of Science:
- Dentistry
- Periodontology
- Radiology
- Artificial Intelligence
- Medical Imaging
Background:
- Accurate detection and classification of periodontal osseous defects are crucial for prognosis and surgical planning in periodontology.
- Traditional radiographic methods have limitations in morphological detail and interpretation consistency.
- Artificial intelligence (AI) shows promise in medical image analysis, but its application to periodontal defect classification on periapical radiographs is underexplored.
Purpose of the Study:
- To evaluate the feasibility of an AI-based object detection model for classifying periodontal osseous defects on periapical radiographs.
- To assess the performance of a YOLOv8 object detection model in identifying and categorizing different types of periodontal defects.
Main Methods:
- Retrospective collection of 7464 periapical radiographs, with 581 annotated images containing periodontal defects.
- Defects were classified into 4 types: 1-wall, 2-or-more-wall, crater-like, and furcation involvements.
- A YOLOv8 large object detection model was trained and evaluated using patient-independent splits, with performance metrics including mAP, precision, and recall.
Main Results:
- The AI model achieved an overall precision of 0.592 and recall of 0.435 (mAP@0.5: 0.504).
- Furcation involvements showed the highest detection rates, while 1-wall defects were the most challenging.
- Smaller or ambiguous defects were more frequently missed, indicating limitations in current AI performance.
Conclusions:
- AI-assisted object detection is feasible for classifying periodontal defects on radiographs but currently has limited performance.
- Dataset imbalance and 2D imaging constraints impact model accuracy.
- Despite limitations, AI models can potentially improve diagnostic consistency and support treatment planning in periodontology.
