POUR-Net: A Population-Prior-Aided Over-Under-Representation Network for Low-Count PET Attenuation Map Generation

Area of Science:

• Medical Imaging
• Radiological Physics
• Artificial Intelligence in Healthcare

Background:

• Low-dose PET imaging minimizes radiation exposure but requires accurate attenuation maps for correction.
• Current methods often use CT scans, increasing overall radiation dose.
• Developing CT-free attenuation map generation is crucial for dose reduction.

Purpose of the Study:

• To propose and evaluate the Population-prior-aided Over-Under-Representation Network (POUR-Net) for generating high-quality attenuation maps from low-dose PET data.
• To enable accurate PET attenuation correction without additional CT scans, thereby reducing radiation exposure.
• To improve the quality of attenuation maps in low-count PET imaging.

Main Methods:

• Developed POUR-Net, integrating an Over-Under-Representation Network (OUR-Net) for efficient feature extraction (low-resolution and fine-detail).
• Incorporated a population prior generation machine (PPGM) using a CT-derived attenuation map dataset to provide prior information.
• Employed a cascade framework for iterative refinement of attenuation map generation.

Main Results:

• POUR-Net successfully generated high-quality attenuation maps from low-dose PET data.
• The method demonstrated effectiveness in CT-free PET attenuation correction.
• Performance surpassed previous baseline methods in experimental evaluations.

Conclusions:

• POUR-Net is a promising solution for accurate, CT-free, low-count PET attenuation correction.
• The proposed method effectively reduces radiation dose in PET imaging.
• This approach advances the field of low-dose PET by enabling high-quality imaging with reduced exposure.