Monte Carlo simulations of pinhole imaging accelerated by kernel-based forced detection

Area of Science:

• Medical Imaging
• Computational Physics
• Nuclear Medicine

Background:

• Pinhole collimation offers superior sensitivity and resolution for small object imaging compared to parallel hole collimation.
• Monte Carlo (MC) simulations are crucial for accurately modeling pinhole imaging but are computationally intensive.
• Accelerating MC simulations is vital for practical applications in pinhole SPECT.

Purpose of the Study:

• To accelerate MC simulations of pinhole SPECT projections.
• To introduce and evaluate a kernel-based forced detection (KFD) technique as an alternative to traditional forced detection (FD).

Main Methods:

• Replaced individual photon tracing in MC simulations with a kernel-based approach (KFD).
• Utilized pre-calculated point spread functions (PSFs) for channel and knife-edge pinhole apertures to model penetration effects.
• Validated KFD speed-up and accuracy using digital phantoms and comparing results with FD simulations.

Main Results:

• KFD and FD simulations produced nearly identical images.
• KFD achieved convergence to equivalent image noise levels one to four orders of magnitude faster than FD.
• The speed-up factor was dependent on the number of photons simulated.

Conclusions:

• KFD is a highly effective technique for accelerating MC simulations in pinhole SPECT.
• The KFD method offers significant computational advantages without compromising image accuracy.
• An MC simulator accelerated by KFD can be a valuable tool for enhancing iterative image reconstruction.