Detection and diagnosis of dental caries using a deep learning-based convolutional neural network algorithm.

Area of Science:

• Artificial intelligence in dentistry
• Medical imaging analysis
• Deep learning for diagnostics

Background:

• Deep convolutional neural networks (CNNs) are emerging in medical research, showing success in radiology and pathology.
• Dental caries detection on radiographs is crucial for oral health management.

Purpose of the Study:

• To evaluate the efficacy of deep CNN algorithms for detecting and diagnosing dental caries.
• To assess the performance of deep CNNs on periapical radiographs.

Main Methods:

• Utilized 3000 periapical radiographs, split into training/validation (80%) and testing (20%) datasets.
• Employed a pre-trained GoogLeNet Inception v3 CNN for preprocessing and transfer learning.
• Calculated diagnostic accuracy, sensitivity, specificity, PPV, NPV, ROC curve, and AUC.

Main Results:

• The deep CNN achieved high diagnostic accuracies: 89.0% for premolars, 88.0% for molars, and 82.0% for both.
• Area Under the Curve (AUC) values were 0.917 (premolars), 0.890 (molars), and 0.845 (both).
• The premolar model demonstrated the highest AUC, significantly outperforming other models (P < 0.001).

Conclusions:

• Deep CNN architecture holds significant potential for dental caries detection and diagnosis.
• The developed deep CNN algorithm demonstrated strong performance in identifying caries on periapical radiographs.
• Deep CNNs are anticipated to become highly effective and efficient tools for dental caries diagnosis.