Evaluation of a Decision Support System Developed with Deep Learning Approach for Detecting Dental Caries with...

Area of Science:

• Dentistry
• Radiology
• Artificial Intelligence

Background:

• Dental caries detection is crucial for timely intervention.
• Cone-beam computed tomography (CBCT) is a valuable imaging tool.
• AI systems offer potential to enhance diagnostic accuracy in radiology.

Purpose of the Study:

• To evaluate the effectiveness of an AI system (Diagnocat) in detecting dental caries on CBCT images.
• To compare the diagnostic performance of dentomaxillofacial radiologists with and without AI assistance.
• To assess the impact of AI on intra- and inter-observer agreement in caries detection.

Main Methods:

• 500 CBCT volumes were analyzed by three radiologists.
• Radiologists scored caries presence with and without AI assistance on a five-point confidence scale.
• Ground truth was established using a hybrid approach; AI used a deep convolutional neural network (CNN).

Main Results:

• AI-aided evaluations significantly improved diagnostic accuracy, with the best accuracy reaching 0.939.
• Area under the ROC curve increased for all observers when using the AI system (e.g., Observer 1: 0.855 to 0.920).
• Fleiss Kappa coefficient improved from 0.325 to 0.468, indicating better inter-observer agreement with AI.

Conclusions:

• The AI system enhances the accuracy and compatibility of radiographic evaluations for dental caries detection using CBCT.
• AI assistance leads to more reliable and consistent diagnoses by radiologists.
• AI tools show significant promise for improving dental diagnostic workflows.