An inhomogeneous most likely path formalism for proton computed tomography

Area of Science:

• Medical Imaging
• Particle Physics
• Computational Science

Background:

• Multiple Coulomb scattering (MCS) complicates proton CT (pCT) image reconstruction.
• Current methods often assume a uniform water-based scattering environment.
• Bayesian models of the most likely path (MLP) are used to address scattering challenges.

Purpose of the Study:

• To develop an MLP formalism for pCT reconstruction that accurately determines scattering moments in inhomogeneous media.
• To improve proton path estimation in pCT by accounting for material composition.
• To enhance the accuracy of pCT image reconstruction in non-water environments.

Main Methods:

• Calculated scattering power relative to water (RScP) for human tissues and compared it with relative stopping power (RStP).
• Utilized Monte Carlo simulations to compare the inhomogeneous MLP formalism with the standard water-based approach.
• Investigated an MLP-Spline-Hybrid method for computational efficiency.

Main Results:

• Established a correlation between RStP and RScP, potentially aiding iterative pCT reconstruction.
• The inhomogeneous formalism accurately predicted proton paths in a water cube with bone inserts (within 1.0 mm).
• Accuracy improvements ranged from 5% to 17% for specific proton energies, though no significant gain was observed in a human head phantom at 200 MeV. The MLP-Spline-Hybrid method halved computation time with minimal accuracy loss.

Conclusions:

• An MLP formalism accounting for material composition in pCT has been developed.
• While a water scattering assumption is often sufficient, the proposed algorithm enhances proton path estimation in significantly heterogeneous tissues.
• The method offers improved accuracy for specific pCT applications involving material variations.