Image-based Modeling of PSF Deformation with Application to Limited Angle PET Data

Area of Science:

• Medical Imaging
• Nuclear Medicine
• Image Reconstruction

Background:

• Point-spread-function (PSF) deformations in positron emission tomography (PET) images arise from detector effects, acquisition geometry, and reconstruction methods.
• These deformations, particularly in dual-panel systems, reduce quantitative accuracy and lesion measurement consistency across the field-of-view (FOV).
• Existing methods may not fully address the complex, spatially variant PSF distortions inherent in advanced PET systems.

Purpose of the Study:

• To develop and evaluate an image-based resolution model (IRM) for correcting spatially variant PSF distortions in dual-panel PET imaging.
• To assess the effectiveness of a spatially variant IRM compared to a spatially invariant model for improving quantitative accuracy and lesion uptake consistency.
• To address reconstruction imperfections caused by limited angles, parallax errors, and other deformation effects in challenging dual-panel PET data.

Main Methods:

• Simulated data from a dedicated breast imaging PET (B-PET) geometry with dual-panel, time-of-flight (TOF) detectors were used.
• Two IRMs were compared: a spatially invariant Gaussian model and a spatially variant Gaussian mixture model (GMM).
• The models were tested for their ability to capture and correct asymmetric PSF shapes and spatially variant deformations.

Main Results:

• Both IRMs reduced overall uptake bias in reconstructed PET images.
• The spatially variant GMM-based IRM effectively ameliorated spatially variant PSF deformation effects.
• This led to more consistent lesion uptake measurements, independent of lesion location within the FOV.

Conclusions:

• Spatially variant image-based resolution modeling is crucial for accurate quantitative analysis in dual-panel PET systems.
• The proposed GMM-based IRM significantly improves lesion uptake consistency and quantitative accuracy.
• This approach enhances the reliability of PET imaging for lesion detection and characterization in challenging geometries.