Monte Carlo-based quantitative pinhole SPECT reconstruction using a ray-tracing back-projector.

Area of Science:

• Nuclear medicine imaging
• Medical physics
• Computational modeling

Background:

• Monte Carlo simulations are crucial for accurate modeling of nuclear medicine imaging systems.
• The accuracy of reconstruction algorithms, like maximum-likelihood-expectation-maximization (ML-EM), is limited by the incorporated models.
• Developing advanced reconstruction methods is essential for improving image quality and quantitative accuracy in SPECT.

Purpose of the Study:

• To develop and evaluate a Monte Carlo-based ML-EM reconstruction method for pinhole SPECT.
• To investigate the quantitative accuracy of a combined Monte Carlo forward-projection and ray-tracing back-projection method.
• To assess the performance of the developed method on a clinical pinhole SPECT camera (GE Discovery NM 530c).

Main Methods:

• Incorporated a Monte Carlo-based model into the SIMIND program for forward-projection, accounting for full physics and geometry.
• Utilized a simplified ray-tracing model for the back-projection step.
• Evaluated the method using 99mTc-filled spheres and a cardiac phantom within a torso phantom.

Main Results:

• The Monte Carlo-based reconstruction generally overestimated total sphere activity but remained within 10% of reference values.
• Activity concentration recovery improved with increasing sphere volume and number of iterations.
• Accurate activity concentration recovery was achieved in cardiac phantom myocardium, irrespective of torso fill status.
• Qualitative comparison showed similar performance to clinical algorithms, with the developed method offering potentially better quantitative accuracy.

Conclusions:

• A Monte Carlo-based reconstruction method for pinhole SPECT was successfully developed and validated with phantom data.
• The combination of Monte Carlo forward-projection and ray-tracing back-projection yielded quantitative images of acceptable quality.
• The method eliminates the need for explicit calibration due to the inherent relationship between counts and activity in the forward-projector model.