Diagnosis and prediction of periodontally compromised teeth using a deep learning-based convolutional neural network...

Area of Science:

• Dentistry
• Artificial Intelligence
• Medical Imaging

Background:

• Periodontally compromised teeth (PCT) pose a diagnostic challenge.
• Accurate diagnosis and prediction of PCT are crucial for treatment planning and prognosis.
• Current diagnostic methods may benefit from advanced computational tools.

Purpose of the Study:

• To develop a computer-assisted detection (CAD) system using a deep convolutional neural network (CNN).
• To evaluate the accuracy and usefulness of the CNN-based system for diagnosing PCT.
• To assess the system's ability to predict the need for tooth extraction in PCT cases.

Main Methods:

• A deep CNN algorithm was developed by combining pre-trained and self-trained networks.
• Periapical radiographic images were utilized for training and validation.
• Diagnostic and predictive performance metrics, including accuracy, sensitivity, specificity, and ROC analysis, were calculated.

Main Results:

• The CNN algorithm achieved diagnostic accuracies of 81.0% for premolars and 76.7% for molars.
• Prediction accuracy for tooth extraction in severe PCT cases was 82.8% for premolars and 73.4% for molars.
• The system demonstrated robust performance on a dataset of 1,740 periapical radiographs.

Conclusions:

• The developed deep CNN algorithm is effective for diagnosing and predicting periodontally compromised teeth (PCT).
• The AI system shows potential as an efficient tool for dental diagnostics.
• Further dataset optimization and algorithm refinement could enhance the CAD system's clinical utility.