Related Experiment Video
Updated: May 31, 2026
07:12
Whole-cell Super-Resolution Imaging via DNA-PAINT on a Spinning Disk Confocal with Optical Photon Reassignment
Published on: January 6, 2026
Online detector response calculations for high-resolution PET image reconstruction
1Department of Radiation Oncology, Stanford University, Stanford, CA 94305, USA.
Physics in Medicine and Biology
|June 17, 2011
Summary
This study introduces a shift-varying model for positron emission tomography (PET) systems, improving image quality and quantitative accuracy in reconstructions. The new model enhances contrast recovery and spatial resolution compared to traditional shift-invariant methods.
Area of Science:
- Medical Imaging
- Nuclear Medicine
- Computational Imaging
Background:
- Positron emission tomography (PET) systems are typically modeled as linear shift-varying (LSV).
- Current image reconstruction often employs simplified shift-invariant (SI) models, compromising image quality and quantitative accuracy.
- Accurate modeling of the geometrical system response is crucial for precise PET imaging.
Purpose of the Study:
- To investigate an analytically formulated shift-varying model for PET geometrical system response.
- To integrate this LSV model into a list-mode, 3D iterative reconstruction algorithm.
- To evaluate the performance of the LSV model against a traditional SI model in terms of image quality and quantitative accuracy.
Main Methods:
- Developed an analytical formulation for a shift-varying system response model.
- Integrated the LSV model into a list-mode, fully 3D iterative reconstruction process.
- Calculated system response coefficients online using a graphics processing unit (GPU) with efficient memory usage (<512 Mb) and processing speed (2 million events/min).
Main Results:
- The analytical LSV model demonstrated good agreement with reference calculations for small detector elements.
- Images reconstructed with the LSV model showed superior quality and quantitative accuracy compared to the SI model.
- Contrast recovery for an 8 mm sphere improved from 85.9% (SI) to 95.8% (LSV).
- Spatial resolution uniformity improved significantly, with Root Mean Square (RMS) variation in reconstructed sphere size reduced from 0.5 mm (SI) to 0.07 mm (LSV) for 1.75 mm spheres.
Conclusions:
- The shift-varying model provides a more accurate representation of PET system geometry than shift-invariant models.
- Implementing an analytical LSV model in GPU-accelerated iterative reconstruction enhances PET image quality and quantitative precision.
- This approach offers significant improvements for clinical PET imaging applications demanding high accuracy.
