SU‐FF‐I‐40

Proton Computed Tomography Reconstruction Using Compressed Sensing and Prior Image Constrained Compressed Sensing

D. X. Wang, T. R. Mackie, Wolfgang A. Tome

Research output: Contribution to journalArticle

1 Citation (Scopus)

Abstract

Purpose: To study the feasibility of proton computed tomography (CT) reconstruction using compressed sensing (CS) and prior image constrained compressed sensing (PICCS). Proton CT images can be used for pre‐therapy planning, image guidance and registration verification. Method and Materials: Projections of 200 MeV proton beams onto an ellipsoid phantom was simulated using Geant4 Monte Carlo simulation toolkit. The position and energy of the entrance and exit protons were recorded. Straight‐line path (SLP) estimation was used to represent proton paths, and simultaneous algebraic reconstruction technique (SART) with CS was used to reconstruct a proton stopping power image for a 2‐mm thick slice of the phantom. PICCS was used to reconstruct the image from highly undersampled data with an accurate and well‐registered prior image. A Gradient transform was used to yield a sparse data set for CS and PICCS. Results: SART with CS reconstructed a 320×320 proton stopping power image of the central slice of the phantom after 10 iterations. A proton/pixel ratio of 0.2 is sufficient to reconstruct an image of correct geometry. The average proton stopping power of the reconstructed materials cortical bone, water, and air were found to agree with the expected values from ICRU Report 49 within 8.3%, 0.6%, and 3.8% respectively. Employing a prior image and PICCS in the reconstruction, a proton/pixel ratio as low as 0.05 was found to be sufficient, and the reconstruction time of less than 2 minutes was achieved using a serial algorithm. Reconstruction artifacts in the images were minimal. Conclusion: With CS, or with PICCS plus a prior image, SART can reconstruct a proton CT image of good quality within minutes. This paves the road to a clinically feasible approach toward low‐dose pre‐proton therapy treatment planning and image guidance using a fast‐reconstructed proton CT image with a well‐registered kV‐CT prior image.

Original languageEnglish (US)
Pages (from-to)2443
Number of pages1
JournalMedical Physics
Volume36
Issue number6
DOIs
StatePublished - 2009
Externally publishedYes

Fingerprint

Protons
Tomography
Feasibility Studies
Artifacts
Air
Water

ASJC Scopus subject areas

  • Biophysics
  • Radiology Nuclear Medicine and imaging

Cite this

SU‐FF‐I‐40 : Proton Computed Tomography Reconstruction Using Compressed Sensing and Prior Image Constrained Compressed Sensing. / Wang, D. X.; Mackie, T. R.; Tome, Wolfgang A.

In: Medical Physics, Vol. 36, No. 6, 2009, p. 2443.

Research output: Contribution to journalArticle

@article{f85add02387a4dd383c51f10e0437396,
title = "SU‐FF‐I‐40: Proton Computed Tomography Reconstruction Using Compressed Sensing and Prior Image Constrained Compressed Sensing",
abstract = "Purpose: To study the feasibility of proton computed tomography (CT) reconstruction using compressed sensing (CS) and prior image constrained compressed sensing (PICCS). Proton CT images can be used for pre‐therapy planning, image guidance and registration verification. Method and Materials: Projections of 200 MeV proton beams onto an ellipsoid phantom was simulated using Geant4 Monte Carlo simulation toolkit. The position and energy of the entrance and exit protons were recorded. Straight‐line path (SLP) estimation was used to represent proton paths, and simultaneous algebraic reconstruction technique (SART) with CS was used to reconstruct a proton stopping power image for a 2‐mm thick slice of the phantom. PICCS was used to reconstruct the image from highly undersampled data with an accurate and well‐registered prior image. A Gradient transform was used to yield a sparse data set for CS and PICCS. Results: SART with CS reconstructed a 320×320 proton stopping power image of the central slice of the phantom after 10 iterations. A proton/pixel ratio of 0.2 is sufficient to reconstruct an image of correct geometry. The average proton stopping power of the reconstructed materials cortical bone, water, and air were found to agree with the expected values from ICRU Report 49 within 8.3{\%}, 0.6{\%}, and 3.8{\%} respectively. Employing a prior image and PICCS in the reconstruction, a proton/pixel ratio as low as 0.05 was found to be sufficient, and the reconstruction time of less than 2 minutes was achieved using a serial algorithm. Reconstruction artifacts in the images were minimal. Conclusion: With CS, or with PICCS plus a prior image, SART can reconstruct a proton CT image of good quality within minutes. This paves the road to a clinically feasible approach toward low‐dose pre‐proton therapy treatment planning and image guidance using a fast‐reconstructed proton CT image with a well‐registered kV‐CT prior image.",
author = "Wang, {D. X.} and Mackie, {T. R.} and Tome, {Wolfgang A.}",
year = "2009",
doi = "10.1118/1.3181159",
language = "English (US)",
volume = "36",
pages = "2443",
journal = "Medical Physics",
issn = "0094-2405",
publisher = "AAPM - American Association of Physicists in Medicine",
number = "6",

}

TY - JOUR

T1 - SU‐FF‐I‐40

T2 - Proton Computed Tomography Reconstruction Using Compressed Sensing and Prior Image Constrained Compressed Sensing

AU - Wang, D. X.

AU - Mackie, T. R.

AU - Tome, Wolfgang A.

PY - 2009

Y1 - 2009

N2 - Purpose: To study the feasibility of proton computed tomography (CT) reconstruction using compressed sensing (CS) and prior image constrained compressed sensing (PICCS). Proton CT images can be used for pre‐therapy planning, image guidance and registration verification. Method and Materials: Projections of 200 MeV proton beams onto an ellipsoid phantom was simulated using Geant4 Monte Carlo simulation toolkit. The position and energy of the entrance and exit protons were recorded. Straight‐line path (SLP) estimation was used to represent proton paths, and simultaneous algebraic reconstruction technique (SART) with CS was used to reconstruct a proton stopping power image for a 2‐mm thick slice of the phantom. PICCS was used to reconstruct the image from highly undersampled data with an accurate and well‐registered prior image. A Gradient transform was used to yield a sparse data set for CS and PICCS. Results: SART with CS reconstructed a 320×320 proton stopping power image of the central slice of the phantom after 10 iterations. A proton/pixel ratio of 0.2 is sufficient to reconstruct an image of correct geometry. The average proton stopping power of the reconstructed materials cortical bone, water, and air were found to agree with the expected values from ICRU Report 49 within 8.3%, 0.6%, and 3.8% respectively. Employing a prior image and PICCS in the reconstruction, a proton/pixel ratio as low as 0.05 was found to be sufficient, and the reconstruction time of less than 2 minutes was achieved using a serial algorithm. Reconstruction artifacts in the images were minimal. Conclusion: With CS, or with PICCS plus a prior image, SART can reconstruct a proton CT image of good quality within minutes. This paves the road to a clinically feasible approach toward low‐dose pre‐proton therapy treatment planning and image guidance using a fast‐reconstructed proton CT image with a well‐registered kV‐CT prior image.

AB - Purpose: To study the feasibility of proton computed tomography (CT) reconstruction using compressed sensing (CS) and prior image constrained compressed sensing (PICCS). Proton CT images can be used for pre‐therapy planning, image guidance and registration verification. Method and Materials: Projections of 200 MeV proton beams onto an ellipsoid phantom was simulated using Geant4 Monte Carlo simulation toolkit. The position and energy of the entrance and exit protons were recorded. Straight‐line path (SLP) estimation was used to represent proton paths, and simultaneous algebraic reconstruction technique (SART) with CS was used to reconstruct a proton stopping power image for a 2‐mm thick slice of the phantom. PICCS was used to reconstruct the image from highly undersampled data with an accurate and well‐registered prior image. A Gradient transform was used to yield a sparse data set for CS and PICCS. Results: SART with CS reconstructed a 320×320 proton stopping power image of the central slice of the phantom after 10 iterations. A proton/pixel ratio of 0.2 is sufficient to reconstruct an image of correct geometry. The average proton stopping power of the reconstructed materials cortical bone, water, and air were found to agree with the expected values from ICRU Report 49 within 8.3%, 0.6%, and 3.8% respectively. Employing a prior image and PICCS in the reconstruction, a proton/pixel ratio as low as 0.05 was found to be sufficient, and the reconstruction time of less than 2 minutes was achieved using a serial algorithm. Reconstruction artifacts in the images were minimal. Conclusion: With CS, or with PICCS plus a prior image, SART can reconstruct a proton CT image of good quality within minutes. This paves the road to a clinically feasible approach toward low‐dose pre‐proton therapy treatment planning and image guidance using a fast‐reconstructed proton CT image with a well‐registered kV‐CT prior image.

UR - http://www.scopus.com/inward/record.url?scp=84874203089&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84874203089&partnerID=8YFLogxK

U2 - 10.1118/1.3181159

DO - 10.1118/1.3181159

M3 - Article

VL - 36

SP - 2443

JO - Medical Physics

JF - Medical Physics

SN - 0094-2405

IS - 6

ER -