A review of PET attenuation correction methods for PET-MR

Area of Science:

• Medical Imaging
• Radiology
• Nuclear Medicine

Background:

• Positron Emission Tomography-Magnetic Resonance (PET-MR) systems are underutilized compared to PET-CT.
• Accurate attenuation correction is a major hurdle for PET-MR clinical implementation.
• Existing MR-based attenuation correction methods show quantitative discrepancies compared to CT- or transmission-based methods.

Purpose of the Study:

• To review and categorize current attenuation correction (AC) approaches for PET-MR.
• To discuss the advantages and disadvantages of each AC category.
• To provide an overview of the current status and future potential of PET-MR AC.

Main Methods:

• Categorization of AC methods into four groups: MR-based, emission-based, atlas-based, and machine learning-based.
• Description and discussion of various techniques within each category.
• Comparative analysis of the outlined AC approaches.

Main Results:

• MR-based AC uses segmented MR images with predefined tissue attenuation coefficients.
• Emission-based AC reconstructs radioactivity and attenuation simultaneously from PET data.
• Atlas-based AC predicts CT/transmission images from MR images using population databases.
• Machine learning, particularly deep learning, is rapidly advancing by learning image features directly from acquired data.

Conclusions:

• Machine learning-based AC, especially deep learning, shows significant promise for accurate PET-MR attenuation correction.
• Further development is needed to overcome quantitative discrepancies and improve reproducibility.
• Addressing these challenges will be crucial for the broader clinical adoption of PET-MR imaging.