Parameter-Transferred Wasserstein Generative Adversarial Network (PT-WGAN) for Low-Dose PET Image Denoising

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Radiology

Background:

• Positron emission tomography (PET) is widely used in clinical settings.
• Reducing radiation dose in PET is essential for patient safety.
• Low-dose PET imaging can result in noisy images, compromising diagnostic accuracy.

Purpose of the Study:

• To develop an effective method for denoising low-dose PET images.
• To preserve structural details and diagnostic performance in low-dose PET scans.
• To improve the efficiency of training deep learning models for PET image enhancement.

Main Methods:

• Proposed a parameter-transferred Wasserstein generative adversarial network (PT-WGAN) for low-dose PET image denoising.
• Implemented a transfer learning strategy for task-specific initialization, utilizing parameters from a pre-trained model.
• Trained and evaluated the PT-WGAN on clinical PET data.

Main Results:

• The PT-WGAN effectively suppressed noise in low-dose PET images.
• The proposed method demonstrated superior image fidelity compared to state-of-the-art techniques.
• Transfer learning significantly enhanced the training efficiency of the PT-WGAN.

Conclusions:

• The PT-WGAN is a promising approach for high-quality low-dose PET imaging.
• The developed method balances noise reduction with the preservation of essential image details.
• The use of transfer learning accelerates the training process for PET image denoising models.