Evaluation of STIR Library Adapted for PET Scanners with Non-Cylindrical Geometry

Area of Science:

• Medical Imaging
• Computational Science

Background:

• The Software for Tomographic Image Reconstruction (STIR) is an open-source C++ library for positron emission tomography (PET) and single photon emission tomography (SPECT) data reconstruction.
• STIR features an experimental scanner geometry modeling capability for precise detector placement.
• Accurate scanner geometry is crucial for high-quality tomographic image reconstruction.

Purpose of the Study:

• To test and enhance the experimental scanner geometry modeling feature in STIR.
• To evaluate the impact of improved geometry modeling on image quality for non-cylindrical PET scanners.
• To ensure compatibility of the revised geometry class with existing STIR features.

Main Methods:

• Utilized Monte Carlo simulations to generate synthetic PET data.
• Employed measured phantom data from a dedicated brain PET prototype scanner.
• Applied the experimental "BlocksOnCylindrical" geometry class to reconstruct PET data.
• Assessed image quality metrics including spatial resolution, uniformity, and contrast.

Main Results:

• The enhanced geometry class significantly improved spatial resolution, uniformity, and image contrast in reconstructed PET images.
• Improvements were particularly evident in the reconstruction of small features within a test quality phantom.
• The revised geometry modeling demonstrated effectiveness for non-cylindrical PET scanner configurations.

Conclusions:

• The "BlocksOnCylindrical" class represents a valuable enhancement for the STIR software library.
• Adjustments to existing STIR features (e.g., Single Scatter Simulation, forward projection, attenuation corrections) are necessary for seamless integration.
• The improved geometry modeling is expected to benefit future PET image reconstruction efforts.