A heuristic statistical stopping rule for iterative reconstruction in emission tomography

Area of Science:

• Medical Imaging
• Nuclear Medicine
• Image Reconstruction

Background:

• Iterative reconstruction is crucial for emission tomography (e.g., PET) but requires careful parameter selection.
• Determining the optimal stopping point for iterative algorithms is challenging and impacts image quality.
• Classical methods often involve post-filtering, which can degrade resolution and introduce artifacts.

Purpose of the Study:

• To introduce a novel statistical stopping criterion for iterative reconstruction in emission tomography.
• To evaluate the performance of this criterion against traditional methods using simulations and realistic data.
• To assess the impact on image quality, specifically the noise-resolution tradeoff.

Main Methods:

• Developed a heuristic statistical stopping criterion based on the reconstruction process.
• Assessed the criterion for Maximum Likelihood Expectation Maximization (MLEM) reconstruction.
• Compared the method against classical iterative reconstruction with low-pass filtering using Monte Carlo simulations and realistic PET data (Hoffman brain phantom).

Main Results:

• The proposed statistical stopping criterion demonstrated significant improvement in the noise-resolution tradeoff compared to classical methods.
• This improvement was observed across a wide range of counting statistics relevant to clinical settings.
• Evaluation with realistic PET data confirmed the criterion's ability to efficiently determine optimal image parameters.

Conclusions:

• The statistical stopping criterion provides a reliable estimation of the optimal stopping point for iterative reconstruction.
• It offers a practical advantage by producing images of comparable or superior quality to post-filtered iterative reconstruction.
• The method achieves this with manageable computational time, making it suitable for clinical application.