Generative adversarial networks in dental imaging: a systematic review

Area of Science:

• Artificial Intelligence in Dentistry
• Medical Image Analysis
• Generative Adversarial Networks (GANs)

Background:

• Generative Adversarial Networks (GANs) are increasingly utilized in medical imaging.
• Dental image analysis presents unique challenges and opportunities for AI-driven solutions.
• This review synthesizes current GAN applications in dental imagery.

Approach:

• Systematic review of electronic databases (PubMed/MEDLINE, Scopus, Embase, Cochrane Library).
• Inclusion of 18 full-text articles on GANs in dental image analysis.
• Assessment of risk of bias and applicability concerns using QUADAS-2 tool.

Key Points:

• GANs applied to 2D and 3D dental images for artifact reduction, denoising, super-resolution, and domain transfer.
• GANs used for image generation, augmentation, outcome prediction, and identification in dental contexts.
• Generated images supported tasks like landmark detection, object detection, and classification.
• Most studies (72%) had low risk of bias, but only 17% had low applicability concerns.

Conclusions:

• GANs demonstrate broad application potential in dental imaging.
• Future research should focus on enhancing GAN stability, repeatability, and interpretability.
• Addressing these limitations will improve GAN applicability and benefit the dental field through AI integration.