General purpose graphics processing unit speedup of integral relative electron density calculation for proton computed tomography

McAllister, Scott; Schubert, Keith E.; Schulte, R.; Penfold, Scott

doi:10.1109/nssmic.2009.5402383

Cited by 3 publications

(3 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The need to estimate proton paths on a one by one basis, coupled with the inability to use many well-established x-ray CT reconstruction methods, comes with a significant computational burden (Johnson 2017). Various avenues of research into overcoming this problem have been explored, from optimising the computer code for MLP evaluation (McAllister et al 2009), to alternative approaches approximating MLP through cubic splines(Collins-Fekete et al 2015) or polynomial approximations (Krah et al 2019).…”

Section: Introductionmentioning

confidence: 99%

Proton path reconstruction for proton computed tomography using neural networks

Ackernley¹,

Casse²,

Cristoforetti³

2021

Phys. Med. Biol.

View full text Add to dashboard Cite

The most likely path formalism (MLP) is widely established as the most statistically precise method for proton path reconstruction in proton computed tomography. However, while this method accounts for small-angle multiple coulomb scattering (MCS) and energy loss, inelastic nuclear interactions play an influential role in a significant number of proton paths. By applying cuts based on energy and direction, tracks influenced by nuclear interactions are largely discarded from the MLP analysis. In this work we propose a new method to estimate the proton paths based on a deep neural network (DNN). Through this approach, estimates of proton paths equivalent to MLP predictions have been achieved in the case where only MCS occurs, together with an increased accuracy when nuclear interactions are present. Moreover, our tests indicate that the DNN algorithm can be considerably faster than the MLP algorithm.

show abstract

Section: Introductionmentioning

confidence: 99%

Proton path reconstruction for proton computed tomography using neural networks

Ackernley¹,

Casse²,

Cristoforetti³

2021

Phys. Med. Biol.

View full text Add to dashboard Cite

show abstract

“…The need to estimate proton paths on a one by one basis, coupled with the inability to use well-established x-ray CT reconstruction methods, comes with a significant computational burden. Various avenues of research into overcoming this problem have been explored, from optimizing the computer code for MLP evaluation (McAllister et al (2009)), to alternative approaches approximating MLP through cubic splines (Fekete et al (2015)) or polynomial approximations (Krah et al (2019)).…”

Section: Introductionmentioning

confidence: 99%

Proton path reconstruction for pCT using Neural Networks

Ackernley,

Casse,

Cristoforetti

2020

Preprint

View full text Add to dashboard Cite

The Most Likely Path formalism (MLP) is widely established as the most statistically precise method for proton path reconstruction in proton computed tomography (pCT). However, while this method accounts for small-angle Multiple Coulomb Scattering (MCS) and energy loss, inelastic nuclear interactions play an influential role in a significant number of proton paths. By applying cuts based on energy and direction, tracks influenced by nuclear interactions are largely discarded from the MLP analysis. In this work we propose a new method to estimate the proton paths based on a Deep Neural Network (DNN). Through this approach, estimates of proton paths equivalent to MLP predictions have been achieved in the case where only MCS occurs, together with an increased accuracy when nuclear interactions are present. Moreover, our tests indicate that the DNN algorithm can be considerably faster than the MLP algorithm.

show abstract

“…For example, only 6 or 8 floating point operations are required per depth u if the MLP is approximated by a polynomial of order M = 0 or M = 1, respectively. On the other hand, evaluating the conventional MLP expression ofSchulte et al (2008) (equation (A.1)) for M = 5 results in at least 47 floating point operations(McAllister 2009), even if sparsity of the matrices R 0 and R 1 is exploited.…”

mentioning

confidence: 99%

Polynomial modelling of proton trajectories in homogeneous media for fast most likely path estimation and trajectory simulation

2019

View full text Add to dashboard Cite

Protons undergo many small angle deflections when traversing a medium, such as a patient. This effect, known as multiple Coulomb scattering (MCS), leads to degraded image resolution in proton radiography and computed tomography (CT) and to lateral spreading of the dose distribution in proton therapy. To optimally account for MCS in proton imaging, the most likely path (MLP) of a proton is estimated based on its position and propagation angle measured in front of and behind the object. In this work, we propose a functional which quantifies the likelihood of a proton trajectory and study how it can be used to model proton trajectories in a homogeneous medium. We focus on two aspects: first, we present an analytical method to quickly generate proton trajectories in a homogeneous medium based on the likelihood functional and validate it through Monte Carlo simulations. It could be used for fast generation of proton CT images without a full Monte Carlo simulation, or potentially to complement the components in a treatment planning Monte Carlo which simulate MCS. Second, by maximising the likelihood functional, we derive an expression for the MLP which is equivalent to the conventional ones reported in the literature yet computationally more convenient. Moreover, we show that the MLP is strictly a polynomial function if the protons' energy loss in the medium is approximated as a polynomial and that the orders of both are linked. We validate our MLP through Monte Carlo simulations and compare proton CT images reconstructed with our expression and with the conventional one. We find that an MLP polynomial of orders larger than five do not lead to increased spatial resolution compared to lower order expressions.

show abstract

General purpose graphics processing unit speedup of integral relative electron density calculation for proton computed tomography

Cited by 3 publications

References 10 publications

Proton path reconstruction for proton computed tomography using neural networks

Proton path reconstruction for proton computed tomography using neural networks

Proton path reconstruction for pCT using Neural Networks

Polynomial modelling of proton trajectories in homogeneous media for fast most likely path estimation and trajectory simulation

Contact Info

Product

Resources

About