On the use of a proton path probability map for proton computed tomography reconstruction

• Published in: Medical physics (Lancaster) 2010-08, Vol.37 (8), p.4138-4145
• Main Authors: Wang, Dongxu, Mackie, T. Rockwell, Tomé, Wolfgang A.
• Format: Article
• Language: English
• Subjects: algebraic reconstruction; Algorithms; Artificial Intelligence; Bayesian inference; Computer Simulation; Data Interpretation, Statistical; image reconstruction; Models, Biological; Models, Statistical; most-likely path estimation; Pattern Recognition, Automated - methods; pCT reconstruction; Protons; Radiographic Image Enhancement - methods; Radiographic Image Interpretation, Computer-Assisted - methods; Reproducibility of Results; Scattering, Radiation; Sensitivity and Specificity
• ISSN: 0094-2405; 2473-4209
• DOI: 10.1118/1.3453767

Purpose: To describe a method to estimate the proton path in proton computed tomography (pCT) reconstruction, which is based on the probability of a proton passing through each point within an object to be imaged. Methods: Based on multiple Coulomb scattering and a semianalytically derived model, the conditional probability of a proton passing through each point within the object given its incoming and exit condition is calculated in a Bayesian inference framework, employing data obtained from Monte Carlo simulation using GEANT4 . The conditional probability at all of the points in the reconstruction plane forms a conditional probability map and can be used for pCT reconstruction. Results: From the generated conditional probability map, a most-likely path (MLP) and a 90% probability envelope around the most-likely path can be extracted and used for pCT reconstruction. The reconstructed pCT image using the conditional probability map yields a smooth pCT image with minor artifacts. pCT reconstructions obtained using the extracted MLP and the 90% probability envelope compare well to reconstructions employing the method of cubic spline proton path estimation. Conclusions: The conditional probability of a proton passing through each point in an object given its entrance and exit condition can be obtained using the proposed method. The extracted MLP and the 90% probability envelope match the proton path recorded in the GEANT4 simulation well. The generated probability map also provides a benchmark for comparing different path estimation methods.