Theoretical study of lesion detectability of MAP reconstruction using computer observers

Area of Science:

• Medical Imaging
• Image Reconstruction
• Quantitative Imaging

Background:

• Low signal-to-noise ratio (SNR) in emission data necessitates advanced image reconstruction techniques.
• Maximum a posteriori (MAP) principle-based statistical methods show promise for improving image quality over conventional filtered backprojection (FBP).

Purpose of the Study:

• To theoretically evaluate lesion detectability using MAP reconstruction with computer observers.
• To compare the performance of MAP reconstruction against FBP reconstruction for positron emission tomography (PET) systems.

Main Methods:

• Theoretical analysis of lesion detectability using computer observers under MAP reconstruction.
• Investigation of different prior models, including quadratic smoothing priors.
• Comparison of prewhitening (PW) and non-prewhitening (NPW) observers.
• Evaluation for both ideal and realistic PET system models.

Main Results:

• For quadratic smoothing priors, prewhitening observer performance in MAP reconstruction is independent of smoothing parameters.
• Non-prewhitening observer performance shows an optimal smoothing point.
• MAP reconstruction yields a higher SNR for lesion detection than FBP in ideal PET systems due to Poisson noise modeling.
• Realistic PET systems further benefit MAP reconstruction through accurate physical photon detection modeling.

Conclusions:

• MAP reconstruction offers superior lesion detectability compared to FBP, particularly in realistic PET scenarios.
• The choice of observer model influences the optimal parameters for MAP reconstruction.
• Theoretical analysis provides a framework for optimizing statistical image reconstruction in emission imaging.