Related Experiment Video
Updated: May 17, 2026
05:49
Reliability of Artificial Intelligence-Based Cone Beam Computed Tomography Integration with Digital Dental Images
Published on: February 23, 2024
A Multi-Task Deep Learning Model for Quality Control of Periapical Radiographs: Simultaneous Technical Error
Wen Fang1,2, Leqi Liu1,2, Xiaokan Wang1,2
1Stomatology Hospital, School of Stomatology, Zhejiang University School of Medicine, Zhejiang Provincial Clinical Research Center for Oral Diseases, Zhejiang Key Laboratory of Oral Biomedical Research, Hangzhou, 310000, China.
Dento Maxillo Facial Radiology
|May 15, 2026
Summary
A new deep learning model automates periapical radiograph quality control by identifying technical errors and grading image quality. This AI system significantly reduces assessment time, enhancing dental workflow efficiency.
Area of Science:
- Artificial Intelligence in Dentistry
- Medical Imaging Analysis
- Radiographic Quality Assurance
Background:
- Automated quality control of periapical radiographs is crucial for accurate dental diagnostics.
- Manual review of radiographs is time-consuming and prone to inter-observer variability.
- Standardized quality grading systems are essential for consistent radiographic assessment.
Purpose of the Study:
- To develop and validate a multi-task deep learning model for automated quality control of periapical radiographs.
- To integrate technical error identification with standardized quality grading within a single model.
- To assess the model's performance and efficiency in a clinical setting.
Main Methods:
- A dataset of 3510 periapical radiographs was used for model development and external validation.
- A multi-task deep learning model was created to classify six technical errors and perform hierarchical quality grading (NRPB and CGDent standards).
- Model performance was evaluated using ROC-AUC, sensitivity, specificity, precision, accuracy, and F1-score, with efficiency gains quantified.
Main Results:
- The EfficientNet-based model demonstrated high performance in identifying technical errors (AUC: 0.842-0.978) and quality grading (AUC: 0.940-0.960 for NRPB, 0.959 for CGDent) on an external test set.
- Computational efficiency was significantly improved, reducing manual review time from over 32 minutes to approximately 1.1 seconds for a typical daily workload.
- Grad-CAM visualizations supported the model's clinical interpretability by aligning its focus with technical errors.
Conclusions:
- The developed multi-task deep learning model effectively automates periapical radiograph quality control.
- The system accurately identifies technical errors and assigns standardized quality grades, streamlining dental workflows.
- This AI-driven approach has substantial potential to enhance quality assurance and efficiency in dental radiology.