Fast accurate iterative reconstruction for low-statistics positron volume imaging

Area of Science:

• Medical Imaging
• Nuclear Medicine
• Image Reconstruction

Background:

• Positron emission tomography (PET) imaging often faces challenges with low-statistics data, particularly in dynamic and whole-body scans.
• Traditional reconstruction methods can lose spatial accuracy when histogramming data into coarse bins, impacting image quality.
• Existing expectation maximization-maximum likelihood (EM-ML) techniques can be computationally intensive and may not fully leverage available data.

Purpose of the Study:

• To develop a Fast Accurate Iterative Reconstruction (FAIR) method for low-statistics PET imaging.
• To improve image resolution, contrast, and noise properties compared to standard EM-ML techniques.
• To optimize reconstruction for list-mode data, preserving sampling accuracy.

Main Methods:

• Developed a FAIR method based on the expectation maximization-maximum likelihood (EM-ML) technique.
• Operated directly on list-mode data, preserving maximum sampling accuracy.
• Implemented a single-pass approach that achieves results comparable to multiple standard ML iterations.

Main Results:

• The FAIR method, using list-mode data, demonstrated improved resolution, contrast, and noise properties.
• Achieved comparable image characteristics to several ML iterations in a single pass.
• Showed suitability for sparse data situations and systems prone to sampling accuracy loss.

Conclusions:

• The FAIR method offers a significant advancement for low-statistics PET imaging by utilizing list-mode data effectively.
• This technique enhances image quality by preserving spatial sampling accuracy and reducing reconstruction time.
• FAIR provides a more efficient and accurate reconstruction solution for challenging PET imaging scenarios.