Automatic Detection and Classification of Dental Restorations in Panoramic Radiographs with Artificial Intelligence Method Developed Using Deep Learning Method: Methodological Studies

The aim of this study is to automatically detect and classify dental restorations in panoramic radiographs using a deep learning-based artificial intelligence (AI) method.

Materials and Methods

Data Preparation

• Data Set: A total of 789 panoramic radiographs from children aged 12-15 years were used. The radiographs were obtained from the radiology archive of Atatürk University Faculty of Dentistry.
• Grouping: Panoramic radiographs were divided into two groups: fillings and root canal treatments.
• Model: The U-Net model, implemented with the PyTorch library, was used. This model was used for the detection and segmentation of restorative materials.

Evaluation of Artificial Intelligence Performance

Evaluation Using Confusion Matrix

• Filling Group: Out of 94 labeled teeth with fillings in 50 images, 89 were true positives, 1 was a false positive, and 4 were false negatives. The sensitivity, precision, and F1 scores were 0.9569, 0.9888, and 0.9726, respectively.
• Root Canal Treatment Group: Out of 76 labeled teeth with root canal treatments in 40 images, 60 were true positives, 0 false positives, and 11 false negatives. The sensitivity, precision, and F1 scores were 0.8450, 1, and 0.9160, respectively.

Results and Discussion

Performance of Deep Learning-Based Artificial Intelligence Models

Deep learning-based AI models performed very well in the automatic detection of restorations in panoramic radiographs obtained from children in the permanent dentition period. AI tools can save clinicians time and assist as a decision support system.

Statistical Analysis

Methods Used to Evaluate Model Performance

A confusion matrix was used to evaluate model performance. This matrix is a table commonly used to describe the performance of a classification model on a set of test data for which the true values are known.

Clinical Relevance of the Study

Use of Artificial Intelligence Applications in Dentistry

Dental clinical applications are open to the innovations brought by technology. The adaptation of AI can be applied in many areas such as caries detection, pre-orthodontic treatment planning, implant planning, and diagnosis of pathologies observed in the jaws. Especially the compatibility of AI with image processing highlights the studies conducted on radiology.

Data Set and Model Training

Detailing the Data Set and Model Training

• Data Set: The study was approved by the Clinical Research Ethics Committee of Atatürk University. The data set was divided into two groups: dental restorations with fillings and root canal treatments.
• Model Training: The model training was performed on a computer with 16 GB RAM and an NVIDIA GeForce GTX 1060Ti graphics card. Before training, each panoramic radiograph was resized from 2,943x1,435 pixels to 1,024x512 pixels.