Artificial intelligence supported single detector multi-energy proton radiography system

Research output: Contribution to journalArticlepeer-review

Standard

Artificial intelligence supported single detector multi-energy proton radiography system. / van der Heyden, Brent; Cohilis, Marie; Souris, Kevin; de Freitas Nascimento, Luana; Sterpin, Edmond.

In: Physics in Medicine and Biology, Vol. 66, 105001, 04.05.2021, p. 1-12.

Research output: Contribution to journalArticlepeer-review

Harvard

Author

van der Heyden, Brent ; Cohilis, Marie ; Souris, Kevin ; de Freitas Nascimento, Luana ; Sterpin, Edmond. / Artificial intelligence supported single detector multi-energy proton radiography system. In: Physics in Medicine and Biology. 2021 ; Vol. 66. pp. 1-12.

Bibtex - Download

@article{e36a3689117f434cb89b6be5e5946bb5,
title = "Artificial intelligence supported single detector multi-energy proton radiography system",
abstract = "Proton radiography imaging was proposed as a promising technique to evaluate internal anatomical changes, to enable pre-treatment patient alignment, and most importantly, to optimize the patient specificCT number to stopping-power ratio conversion. The clinical implementation rate of proton radiography systems is still limited due to their complex bulky design, together with the persistent problem of (in)elastic nuclear interactions and multiple Coulomb scattering (i.e. range mixing). In this work, a compact multi-energy proton radiography system was proposed in combination with an artificial intelligence network architecture (ProtonDSE) to remove the persistent problem of proton scatter in proton radiography. A realistic Monte Carlo model of the Proteus{\textregistered}One accelerator was built at 200 and 220MeV to isolate the scattered proton signal in 236 proton radiographies of 80 digital anthropomorphic phantoms. ProtonDSE was trained to predict the proton scatter distribution at two beam energies in a 60%/25%/15% scheme for training, testing, and validation. A calibration procedure was proposed to derive the water equivalent thickness image based on the detector dose response relationship at both beam energies. ProtonDSE network performance was evaluated with quantitative metrics that showed an overall mean absolute percentage error below 1.4%.±.0.4% in our test dataset. For one example patient, detector dose toWETconversions were performed based on the total dose (ITotal), the primary proton dose (IPrimary), and the ProtonDSE corrected detector dose (ICorrected). The determinedWETaccuracy was compared with respect to the referenceWETby idealistic raytracing in a manually delineated region-of-interest inside the brain. The error was determined 4.3%.±.4.1%for WET(ITotal), 2.2%.±.1.4%for WET(IPrimary), and 2.5%.±.2.0% for WET(ICorrected).",
keywords = "Proton radiography, Artificial intelligence, Deep learning, ProtonDSE",
author = "{van der Heyden}, Brent and Marie Cohilis and Kevin Souris and {de Freitas Nascimento}, Luana and Edmond Sterpin",
note = "Score=10",
year = "2021",
month = may,
day = "4",
doi = "10.1088/1361-6560/abe918",
language = "English",
volume = "66",
pages = "1--12",
journal = "Physics in Medicine and Biology",
issn = "0031-9155",
publisher = "IOP - IOP Publishing",

}

RIS - Download

TY - JOUR

T1 - Artificial intelligence supported single detector multi-energy proton radiography system

AU - van der Heyden, Brent

AU - Cohilis, Marie

AU - Souris, Kevin

AU - de Freitas Nascimento, Luana

AU - Sterpin, Edmond

N1 - Score=10

PY - 2021/5/4

Y1 - 2021/5/4

N2 - Proton radiography imaging was proposed as a promising technique to evaluate internal anatomical changes, to enable pre-treatment patient alignment, and most importantly, to optimize the patient specificCT number to stopping-power ratio conversion. The clinical implementation rate of proton radiography systems is still limited due to their complex bulky design, together with the persistent problem of (in)elastic nuclear interactions and multiple Coulomb scattering (i.e. range mixing). In this work, a compact multi-energy proton radiography system was proposed in combination with an artificial intelligence network architecture (ProtonDSE) to remove the persistent problem of proton scatter in proton radiography. A realistic Monte Carlo model of the Proteus®One accelerator was built at 200 and 220MeV to isolate the scattered proton signal in 236 proton radiographies of 80 digital anthropomorphic phantoms. ProtonDSE was trained to predict the proton scatter distribution at two beam energies in a 60%/25%/15% scheme for training, testing, and validation. A calibration procedure was proposed to derive the water equivalent thickness image based on the detector dose response relationship at both beam energies. ProtonDSE network performance was evaluated with quantitative metrics that showed an overall mean absolute percentage error below 1.4%.±.0.4% in our test dataset. For one example patient, detector dose toWETconversions were performed based on the total dose (ITotal), the primary proton dose (IPrimary), and the ProtonDSE corrected detector dose (ICorrected). The determinedWETaccuracy was compared with respect to the referenceWETby idealistic raytracing in a manually delineated region-of-interest inside the brain. The error was determined 4.3%.±.4.1%for WET(ITotal), 2.2%.±.1.4%for WET(IPrimary), and 2.5%.±.2.0% for WET(ICorrected).

AB - Proton radiography imaging was proposed as a promising technique to evaluate internal anatomical changes, to enable pre-treatment patient alignment, and most importantly, to optimize the patient specificCT number to stopping-power ratio conversion. The clinical implementation rate of proton radiography systems is still limited due to their complex bulky design, together with the persistent problem of (in)elastic nuclear interactions and multiple Coulomb scattering (i.e. range mixing). In this work, a compact multi-energy proton radiography system was proposed in combination with an artificial intelligence network architecture (ProtonDSE) to remove the persistent problem of proton scatter in proton radiography. A realistic Monte Carlo model of the Proteus®One accelerator was built at 200 and 220MeV to isolate the scattered proton signal in 236 proton radiographies of 80 digital anthropomorphic phantoms. ProtonDSE was trained to predict the proton scatter distribution at two beam energies in a 60%/25%/15% scheme for training, testing, and validation. A calibration procedure was proposed to derive the water equivalent thickness image based on the detector dose response relationship at both beam energies. ProtonDSE network performance was evaluated with quantitative metrics that showed an overall mean absolute percentage error below 1.4%.±.0.4% in our test dataset. For one example patient, detector dose toWETconversions were performed based on the total dose (ITotal), the primary proton dose (IPrimary), and the ProtonDSE corrected detector dose (ICorrected). The determinedWETaccuracy was compared with respect to the referenceWETby idealistic raytracing in a manually delineated region-of-interest inside the brain. The error was determined 4.3%.±.4.1%for WET(ITotal), 2.2%.±.1.4%for WET(IPrimary), and 2.5%.±.2.0% for WET(ICorrected).

KW - Proton radiography

KW - Artificial intelligence

KW - Deep learning

KW - ProtonDSE

UR - https://ecm.sckcen.be/OTCS/llisapi.dll/open/48096684

U2 - 10.1088/1361-6560/abe918

DO - 10.1088/1361-6560/abe918

M3 - Article

VL - 66

SP - 1

EP - 12

JO - Physics in Medicine and Biology

JF - Physics in Medicine and Biology

SN - 0031-9155

M1 - 105001

ER -

ID: 7410758