Volume assessment accuracy in computed tomography: A phantom study

Nicolas D. Prionas, Shonket Ray, John M Boone

Research output: Contribution to journalArticle

35 Citations (Scopus)

Abstract

There is a broad push in the cancer imaging community to eventually replace linear tumor measurements with three-dimensional evaluation of tumor volume. To evaluate the potential accuracy of volume measurement in tumors by CT, a gelatin phantom consisting of 55 polymethylmethacrylate (PMMA) spheres spanning diameters from 1.6 mm to 25.4 mm was fabricated and scanned using thin slice (0.625 mm) CT (GE LightSpeed 16). Nine different reconstruction combinations of field of view dimension (FOV = 20, 30, 40 cm) and CT kernel (standard, lung, bone) were analyzed. Contiguous thin-slice images were averaged to produce CT images with greater thicknesses (1.25, 2.50, 5.0 mm). Simple grayscale thresholding techniques were used to segment the PMMA spheres from the gelatin background, where a total of 1800 spherical volumes were evaluated across the permutations studied. The geometric simplicity of the phantom established upper limits on measurement accuracy. In general, smaller slice thickness and larger sphere diameters produced more accurate volume assessment than larger slice thickness and smaller sphere diameter. The measured volumes were smaller than the actual volumes by a common factor depending on slice thickness; overall, 0.625 mm slices produced on average 18%, 1.25 mm slices produced 22%, 2.5 mm CT slices produced 29%, and 5.0 mm slices produced 39% underestimates of volume (mm3). Field of view did not have a significant effect on volume accuracy. Reconstruction algorithm significantly affected volume accuracy (p < 0.0001), with the lung kernel having the smallest error, followed by the bone and standard kernels. The results of this investigation provide guidance for CT protocol development and may guide the development of more advanced techniques to promote quantitatively accurate CT volumetric analysis of tumors.

Original languageEnglish (US)
Pages (from-to)168-180
Number of pages13
JournalJournal of Applied Clinical Medical Physics
Volume11
Issue number2
StatePublished - 2010

Fingerprint

Tomography
Tumors
tomography
Polymethyl Methacrylate
Gelatin
Neoplasms
Bone
Volumetric analysis
tumors
Bone and Bones
Volume measurement
Lung
Cone-Beam Computed Tomography
Tumor Burden
gelatins
lungs
bones
field of view
volumetric analysis
Imaging techniques

Keywords

  • Accuracy
  • Computed tomography
  • Phantom
  • Volumetric

ASJC Scopus subject areas

  • Radiology Nuclear Medicine and imaging
  • Radiation
  • Instrumentation

Cite this

Volume assessment accuracy in computed tomography : A phantom study. / Prionas, Nicolas D.; Ray, Shonket; Boone, John M.

In: Journal of Applied Clinical Medical Physics, Vol. 11, No. 2, 2010, p. 168-180.

Research output: Contribution to journalArticle

@article{b9ff5a28c88b4da8b822ccc3121abbd0,
title = "Volume assessment accuracy in computed tomography: A phantom study",
abstract = "There is a broad push in the cancer imaging community to eventually replace linear tumor measurements with three-dimensional evaluation of tumor volume. To evaluate the potential accuracy of volume measurement in tumors by CT, a gelatin phantom consisting of 55 polymethylmethacrylate (PMMA) spheres spanning diameters from 1.6 mm to 25.4 mm was fabricated and scanned using thin slice (0.625 mm) CT (GE LightSpeed 16). Nine different reconstruction combinations of field of view dimension (FOV = 20, 30, 40 cm) and CT kernel (standard, lung, bone) were analyzed. Contiguous thin-slice images were averaged to produce CT images with greater thicknesses (1.25, 2.50, 5.0 mm). Simple grayscale thresholding techniques were used to segment the PMMA spheres from the gelatin background, where a total of 1800 spherical volumes were evaluated across the permutations studied. The geometric simplicity of the phantom established upper limits on measurement accuracy. In general, smaller slice thickness and larger sphere diameters produced more accurate volume assessment than larger slice thickness and smaller sphere diameter. The measured volumes were smaller than the actual volumes by a common factor depending on slice thickness; overall, 0.625 mm slices produced on average 18{\%}, 1.25 mm slices produced 22{\%}, 2.5 mm CT slices produced 29{\%}, and 5.0 mm slices produced 39{\%} underestimates of volume (mm3). Field of view did not have a significant effect on volume accuracy. Reconstruction algorithm significantly affected volume accuracy (p < 0.0001), with the lung kernel having the smallest error, followed by the bone and standard kernels. The results of this investigation provide guidance for CT protocol development and may guide the development of more advanced techniques to promote quantitatively accurate CT volumetric analysis of tumors.",
keywords = "Accuracy, Computed tomography, Phantom, Volumetric",
author = "Prionas, {Nicolas D.} and Shonket Ray and Boone, {John M}",
year = "2010",
language = "English (US)",
volume = "11",
pages = "168--180",
journal = "Journal of Applied Clinical Medical Physics",
issn = "1526-9914",
publisher = "American Institute of Physics Publising LLC",
number = "2",

}

TY - JOUR

T1 - Volume assessment accuracy in computed tomography

T2 - A phantom study

AU - Prionas, Nicolas D.

AU - Ray, Shonket

AU - Boone, John M

PY - 2010

Y1 - 2010

N2 - There is a broad push in the cancer imaging community to eventually replace linear tumor measurements with three-dimensional evaluation of tumor volume. To evaluate the potential accuracy of volume measurement in tumors by CT, a gelatin phantom consisting of 55 polymethylmethacrylate (PMMA) spheres spanning diameters from 1.6 mm to 25.4 mm was fabricated and scanned using thin slice (0.625 mm) CT (GE LightSpeed 16). Nine different reconstruction combinations of field of view dimension (FOV = 20, 30, 40 cm) and CT kernel (standard, lung, bone) were analyzed. Contiguous thin-slice images were averaged to produce CT images with greater thicknesses (1.25, 2.50, 5.0 mm). Simple grayscale thresholding techniques were used to segment the PMMA spheres from the gelatin background, where a total of 1800 spherical volumes were evaluated across the permutations studied. The geometric simplicity of the phantom established upper limits on measurement accuracy. In general, smaller slice thickness and larger sphere diameters produced more accurate volume assessment than larger slice thickness and smaller sphere diameter. The measured volumes were smaller than the actual volumes by a common factor depending on slice thickness; overall, 0.625 mm slices produced on average 18%, 1.25 mm slices produced 22%, 2.5 mm CT slices produced 29%, and 5.0 mm slices produced 39% underestimates of volume (mm3). Field of view did not have a significant effect on volume accuracy. Reconstruction algorithm significantly affected volume accuracy (p < 0.0001), with the lung kernel having the smallest error, followed by the bone and standard kernels. The results of this investigation provide guidance for CT protocol development and may guide the development of more advanced techniques to promote quantitatively accurate CT volumetric analysis of tumors.

AB - There is a broad push in the cancer imaging community to eventually replace linear tumor measurements with three-dimensional evaluation of tumor volume. To evaluate the potential accuracy of volume measurement in tumors by CT, a gelatin phantom consisting of 55 polymethylmethacrylate (PMMA) spheres spanning diameters from 1.6 mm to 25.4 mm was fabricated and scanned using thin slice (0.625 mm) CT (GE LightSpeed 16). Nine different reconstruction combinations of field of view dimension (FOV = 20, 30, 40 cm) and CT kernel (standard, lung, bone) were analyzed. Contiguous thin-slice images were averaged to produce CT images with greater thicknesses (1.25, 2.50, 5.0 mm). Simple grayscale thresholding techniques were used to segment the PMMA spheres from the gelatin background, where a total of 1800 spherical volumes were evaluated across the permutations studied. The geometric simplicity of the phantom established upper limits on measurement accuracy. In general, smaller slice thickness and larger sphere diameters produced more accurate volume assessment than larger slice thickness and smaller sphere diameter. The measured volumes were smaller than the actual volumes by a common factor depending on slice thickness; overall, 0.625 mm slices produced on average 18%, 1.25 mm slices produced 22%, 2.5 mm CT slices produced 29%, and 5.0 mm slices produced 39% underestimates of volume (mm3). Field of view did not have a significant effect on volume accuracy. Reconstruction algorithm significantly affected volume accuracy (p < 0.0001), with the lung kernel having the smallest error, followed by the bone and standard kernels. The results of this investigation provide guidance for CT protocol development and may guide the development of more advanced techniques to promote quantitatively accurate CT volumetric analysis of tumors.

KW - Accuracy

KW - Computed tomography

KW - Phantom

KW - Volumetric

UR - http://www.scopus.com/inward/record.url?scp=77954367861&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=77954367861&partnerID=8YFLogxK

M3 - Article

C2 - 20592693

AN - SCOPUS:77954367861

VL - 11

SP - 168

EP - 180

JO - Journal of Applied Clinical Medical Physics

JF - Journal of Applied Clinical Medical Physics

SN - 1526-9914

IS - 2

ER -