A review on AI in PET imaging

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Radiology

Background:

• Artificial intelligence (AI) is increasingly used in medical imaging for tasks like computer-aided diagnosis.
• Deep learning techniques, specifically CNNs and GANs, are prominent in medical image generation.
• Previous research has explored deep learning for generating images using positron emission tomography (PET).

Purpose of the Study:

• To review studies applying deep learning techniques for image generation in PET.
• To categorize existing research into key themes of PET image generation using deep learning.
• To discuss limitations and future directions for deep learning in PET image generation.

Main Methods:

• Literature review of studies utilizing deep learning for PET image generation.
• Categorization of reviewed studies into three main themes: denoising, reconstruction/attenuation correction, and image translation/synthesis.
• Analysis of recent advancements and methodologies within each category.

Main Results:

• Deep learning effectively aids in denoising noisy PET data to recover full information.
• AI techniques show promise in PET image reconstruction and attenuation correction.
• Deep learning facilitates PET image translation and synthesis, expanding imaging capabilities.

Conclusions:

• Deep learning offers significant potential for advancing PET image generation across various applications.
• Further research is needed to address the limitations and fully realize the future prospects of AI in PET imaging.
• The reviewed categories highlight the diverse and evolving role of deep learning in PET.