A Review on Low-Dose Emission Tomography Post-Reconstruction Denoising with Neural Network Approaches

Area of Science:

• Medical Imaging
• Nuclear Medicine
• Computational Imaging

Background:

• Low-dose emission tomography (ET) is vital for functional medical imaging, but photon counting randomness amplifies noise.
• Image quality degradation in low-dose ET hinders accurate diagnosis and research.

Purpose of the Study:

• To review current post-processing techniques for low-dose ET.
• To emphasize the role of deep neural networks (NNs) in noise reduction and image enhancement.
• To explore future advancements in NN-based low-dose ET.

Main Methods:

• Literature review of post-processing methods for low-dose ET.
• Focus on deep learning and neural network (NN) applications.
• Analysis of current research and future trends in NN-based ET.

Main Results:

• Deep neural networks show significant potential in mitigating noise in low-dose ET.
• NNs can enhance image quality and resolution, improving diagnostic accuracy.
• Various NN architectures are being explored for optimal performance.

Conclusions:

• Deep learning offers a powerful approach to overcome noise limitations in low-dose ET.
• Advancements in NN-based techniques promise to significantly improve medical imaging capabilities.
• Further research into NN architectures and training strategies is crucial for future development.