Generative Adversarial Network (Generative Artificial Intelligence) in Pediatric Radiology: A Systematic Review.

Area of Science:

• Radiology
• Artificial Intelligence
• Pediatric Imaging

Background:

• Generative artificial intelligence (AI), particularly generative adversarial networks (GANs), is a rapidly evolving field in medical imaging.
• While GANs have been reviewed in general radiology, a specific review for pediatric radiology is lacking.

Purpose of the Study:

• To systematically review the applications of GANs in pediatric radiology.
• To evaluate the performance of GAN models and the methods used for their assessment.

Main Methods:

• A systematic literature search was conducted on electronic databases up to April 6, 2023.
• Thirty-seven relevant papers were included in the review.

Main Results:

• GANs are applied across various pediatric imaging modalities (MRI, X-ray, CT, ultrasound, PET) for tasks including image translation, segmentation, reconstruction, quality assessment, synthesis, data augmentation, and disease diagnosis.
• Approximately 80% of studies reported GAN model performance superior to other methods, with improvements ranging from 0.1% to 158.6%.
• Methodological weaknesses were identified in the reviewed studies.

Conclusions:

• GANs offer diverse applications and potential performance benefits in pediatric radiology.
• Methodological limitations in current research may hinder the clinical adoption and realization of GANs' full potential in this field.
• Future research should focus on developing more robust methodologies to address identified weaknesses.