Diagnostic charting of panoramic radiography using deep-learning artificial intelligence system

Development of AI Model CranioCatch

The AI model CranioCatch was developed based on a deep Convolutional Neural Network (CNN) method. The model utilized the Faster R-CNN architecture, which includes a Region Proposal Network (RPN) for efficient region of interest (ROI) detection. The implementation of RPN after the last convolution layer of CNN allowed for accurate and fast detection of dental situations on panoramic radiographs.

Labeling of Dental Situations

Two experienced dento-maxillofacial radiologists manually labeled ground truth images for ten different dental situations on panoramic radiographs. These situations included crown, pontic, root-canal treated tooth, implant, implant-supported crown, impacted tooth, residual root, filling, caries, and dental calculus. The labeling process ensured consensus between the radiologists and achieved high Intraobserver agreement.

Significance of Deep-Learning AI in Diagnostic Charting

The study aimed to develop AI algorithms that could assist in routine clinical examinations by accurately detecting various dental conditions on panoramic radiographs. By utilizing deep-learning AI models like CranioCatch, the goal was to create a decision support system that could aid physicians in diagnosis and treatment planning. The use of AI in dental radiology holds promise for improving diagnostic accuracy, efficiency, and potentially enhancing digital-based student education.

Future Implications and Conclusion

The successful development and evaluation of the CranioCatch AI model suggest a potential shift towards AI-assisted diagnostic charting in dental radiology. Further research with larger study samples is recommended to enhance the success rate of AI models. The authors predict that AI models will play a crucial role in panoramic diagnosis, treatment planning, and dental education, especially in the context of online training during the current pandemic period.