Three-dimensional lung tumor segmentation from x-ray computed tomography using sparse field active models

Joseph Awad, Amir Owrangi, Lauren Villemaire, Elaine O'Riordan, Grace Parraga, Aaron Fenster

Research output: Contribution to journalArticle

15 Citations (Scopus)

Abstract

Purpose: Manual segmentation of lung tumors is observer dependent and time-consuming but an important component of radiology and radiation oncology workflow. The objective of this study was to generate an automated lung tumor measurement tool for segmentation of pulmonary metastatic tumors from x-ray computed tomography (CT) images to improve reproducibility and decrease the time required to segment tumor boundaries. Methods: The authors developed an automated lung tumor segmentation algorithm for volumetric image analysis of chest CT images using shape constrained Otsu multithresholding (SCOMT) and sparse field active surface (SFAS) algorithms. The observer was required to select the tumor center and the SCOMT algorithm subsequently created an initial surface that was deformed using level set SFAS to minimize the total energy consisting of mean separation, edge, partial volume, rolling, distribution, background, shape, volume, smoothness, and curvature energies. Results: The proposed segmentation algorithm was compared to manual segmentation whereby 21 tumors were evaluated using one-dimensional (1D) response evaluation criteria in solid tumors (RECIST), two-dimensional (2D) World Health Organization (WHO), and 3D volume measurements. Linear regression goodness-of-fit measures (r 2 0.63, p 0.0001; r2 0.87, p 0.0001; and r2 0.96, p 0.0001), and Pearson correlation coefficients (r 0.79, p 0.0001; r 0.93, p 0.0001; and r 0.98, p 0.0001) for 1D, 2D, and 3D measurements, respectively, showed significant correlations between manual and algorithm results. Intra-observer intraclass correlation coefficients (ICC) demonstrated high reproducibility for algorithm (0.989-0.995, 0.996-0.997, and 0.999-0.999) and manual measurements (0.975-0.993, 0.985-0.993, and 0.980-0.992) for 1D, 2D, and 3D measurements, respectively. The intra-observer coefficient of variation (CV) was low for algorithm (3.09-4.67, 4.85-5.84, and 5.65-5.88) and manual observers (4.20-6.61, 8.14-9.57, and 14.57-21.61) for 1D, 2D, and 3D measurements, respectively. Conclusions: The authors developed an automated segmentation algorithm requiring only that the operator select the tumor to measure pulmonary metastatic tumors in 1D, 2D, and 3D. Algorithm and manual measurements were significantly correlated. Since the algorithm segmentation involves selection of a single seed point, it resulted in reduced intra-observer variability and decreased time, for making the measurements.

Original languageEnglish (US)
Pages (from-to)851-865
Number of pages15
JournalMedical Physics
Volume39
Issue number2
DOIs
StatePublished - Jan 1 2012

Fingerprint

Tomography
X-Rays
Lung
Neoplasms
Observer Variation
Radiation Oncology
Workflow
Radiology
Linear Models
Seeds
Thorax

Keywords

  • 3D segmentation
  • CT images
  • level set
  • Lung tumors
  • sparse field active surface

ASJC Scopus subject areas

  • Biophysics
  • Radiology Nuclear Medicine and imaging

Cite this

Three-dimensional lung tumor segmentation from x-ray computed tomography using sparse field active models. / Awad, Joseph; Owrangi, Amir; Villemaire, Lauren; O'Riordan, Elaine; Parraga, Grace; Fenster, Aaron.

In: Medical Physics, Vol. 39, No. 2, 01.01.2012, p. 851-865.

Research output: Contribution to journalArticle

Awad, Joseph ; Owrangi, Amir ; Villemaire, Lauren ; O'Riordan, Elaine ; Parraga, Grace ; Fenster, Aaron. / Three-dimensional lung tumor segmentation from x-ray computed tomography using sparse field active models. In: Medical Physics. 2012 ; Vol. 39, No. 2. pp. 851-865.
@article{015ae2b57fde4574a14ed7b30b773dfb,
title = "Three-dimensional lung tumor segmentation from x-ray computed tomography using sparse field active models",
abstract = "Purpose: Manual segmentation of lung tumors is observer dependent and time-consuming but an important component of radiology and radiation oncology workflow. The objective of this study was to generate an automated lung tumor measurement tool for segmentation of pulmonary metastatic tumors from x-ray computed tomography (CT) images to improve reproducibility and decrease the time required to segment tumor boundaries. Methods: The authors developed an automated lung tumor segmentation algorithm for volumetric image analysis of chest CT images using shape constrained Otsu multithresholding (SCOMT) and sparse field active surface (SFAS) algorithms. The observer was required to select the tumor center and the SCOMT algorithm subsequently created an initial surface that was deformed using level set SFAS to minimize the total energy consisting of mean separation, edge, partial volume, rolling, distribution, background, shape, volume, smoothness, and curvature energies. Results: The proposed segmentation algorithm was compared to manual segmentation whereby 21 tumors were evaluated using one-dimensional (1D) response evaluation criteria in solid tumors (RECIST), two-dimensional (2D) World Health Organization (WHO), and 3D volume measurements. Linear regression goodness-of-fit measures (r 2 0.63, p 0.0001; r2 0.87, p 0.0001; and r2 0.96, p 0.0001), and Pearson correlation coefficients (r 0.79, p 0.0001; r 0.93, p 0.0001; and r 0.98, p 0.0001) for 1D, 2D, and 3D measurements, respectively, showed significant correlations between manual and algorithm results. Intra-observer intraclass correlation coefficients (ICC) demonstrated high reproducibility for algorithm (0.989-0.995, 0.996-0.997, and 0.999-0.999) and manual measurements (0.975-0.993, 0.985-0.993, and 0.980-0.992) for 1D, 2D, and 3D measurements, respectively. The intra-observer coefficient of variation (CV) was low for algorithm (3.09-4.67, 4.85-5.84, and 5.65-5.88) and manual observers (4.20-6.61, 8.14-9.57, and 14.57-21.61) for 1D, 2D, and 3D measurements, respectively. Conclusions: The authors developed an automated segmentation algorithm requiring only that the operator select the tumor to measure pulmonary metastatic tumors in 1D, 2D, and 3D. Algorithm and manual measurements were significantly correlated. Since the algorithm segmentation involves selection of a single seed point, it resulted in reduced intra-observer variability and decreased time, for making the measurements.",
keywords = "3D segmentation, CT images, level set, Lung tumors, sparse field active surface",
author = "Joseph Awad and Amir Owrangi and Lauren Villemaire and Elaine O'Riordan and Grace Parraga and Aaron Fenster",
year = "2012",
month = "1",
day = "1",
doi = "10.1118/1.3676687",
language = "English (US)",
volume = "39",
pages = "851--865",
journal = "Medical Physics",
issn = "0094-2405",
publisher = "AAPM - American Association of Physicists in Medicine",
number = "2",

}

TY - JOUR

T1 - Three-dimensional lung tumor segmentation from x-ray computed tomography using sparse field active models

AU - Awad, Joseph

AU - Owrangi, Amir

AU - Villemaire, Lauren

AU - O'Riordan, Elaine

AU - Parraga, Grace

AU - Fenster, Aaron

PY - 2012/1/1

Y1 - 2012/1/1

N2 - Purpose: Manual segmentation of lung tumors is observer dependent and time-consuming but an important component of radiology and radiation oncology workflow. The objective of this study was to generate an automated lung tumor measurement tool for segmentation of pulmonary metastatic tumors from x-ray computed tomography (CT) images to improve reproducibility and decrease the time required to segment tumor boundaries. Methods: The authors developed an automated lung tumor segmentation algorithm for volumetric image analysis of chest CT images using shape constrained Otsu multithresholding (SCOMT) and sparse field active surface (SFAS) algorithms. The observer was required to select the tumor center and the SCOMT algorithm subsequently created an initial surface that was deformed using level set SFAS to minimize the total energy consisting of mean separation, edge, partial volume, rolling, distribution, background, shape, volume, smoothness, and curvature energies. Results: The proposed segmentation algorithm was compared to manual segmentation whereby 21 tumors were evaluated using one-dimensional (1D) response evaluation criteria in solid tumors (RECIST), two-dimensional (2D) World Health Organization (WHO), and 3D volume measurements. Linear regression goodness-of-fit measures (r 2 0.63, p 0.0001; r2 0.87, p 0.0001; and r2 0.96, p 0.0001), and Pearson correlation coefficients (r 0.79, p 0.0001; r 0.93, p 0.0001; and r 0.98, p 0.0001) for 1D, 2D, and 3D measurements, respectively, showed significant correlations between manual and algorithm results. Intra-observer intraclass correlation coefficients (ICC) demonstrated high reproducibility for algorithm (0.989-0.995, 0.996-0.997, and 0.999-0.999) and manual measurements (0.975-0.993, 0.985-0.993, and 0.980-0.992) for 1D, 2D, and 3D measurements, respectively. The intra-observer coefficient of variation (CV) was low for algorithm (3.09-4.67, 4.85-5.84, and 5.65-5.88) and manual observers (4.20-6.61, 8.14-9.57, and 14.57-21.61) for 1D, 2D, and 3D measurements, respectively. Conclusions: The authors developed an automated segmentation algorithm requiring only that the operator select the tumor to measure pulmonary metastatic tumors in 1D, 2D, and 3D. Algorithm and manual measurements were significantly correlated. Since the algorithm segmentation involves selection of a single seed point, it resulted in reduced intra-observer variability and decreased time, for making the measurements.

AB - Purpose: Manual segmentation of lung tumors is observer dependent and time-consuming but an important component of radiology and radiation oncology workflow. The objective of this study was to generate an automated lung tumor measurement tool for segmentation of pulmonary metastatic tumors from x-ray computed tomography (CT) images to improve reproducibility and decrease the time required to segment tumor boundaries. Methods: The authors developed an automated lung tumor segmentation algorithm for volumetric image analysis of chest CT images using shape constrained Otsu multithresholding (SCOMT) and sparse field active surface (SFAS) algorithms. The observer was required to select the tumor center and the SCOMT algorithm subsequently created an initial surface that was deformed using level set SFAS to minimize the total energy consisting of mean separation, edge, partial volume, rolling, distribution, background, shape, volume, smoothness, and curvature energies. Results: The proposed segmentation algorithm was compared to manual segmentation whereby 21 tumors were evaluated using one-dimensional (1D) response evaluation criteria in solid tumors (RECIST), two-dimensional (2D) World Health Organization (WHO), and 3D volume measurements. Linear regression goodness-of-fit measures (r 2 0.63, p 0.0001; r2 0.87, p 0.0001; and r2 0.96, p 0.0001), and Pearson correlation coefficients (r 0.79, p 0.0001; r 0.93, p 0.0001; and r 0.98, p 0.0001) for 1D, 2D, and 3D measurements, respectively, showed significant correlations between manual and algorithm results. Intra-observer intraclass correlation coefficients (ICC) demonstrated high reproducibility for algorithm (0.989-0.995, 0.996-0.997, and 0.999-0.999) and manual measurements (0.975-0.993, 0.985-0.993, and 0.980-0.992) for 1D, 2D, and 3D measurements, respectively. The intra-observer coefficient of variation (CV) was low for algorithm (3.09-4.67, 4.85-5.84, and 5.65-5.88) and manual observers (4.20-6.61, 8.14-9.57, and 14.57-21.61) for 1D, 2D, and 3D measurements, respectively. Conclusions: The authors developed an automated segmentation algorithm requiring only that the operator select the tumor to measure pulmonary metastatic tumors in 1D, 2D, and 3D. Algorithm and manual measurements were significantly correlated. Since the algorithm segmentation involves selection of a single seed point, it resulted in reduced intra-observer variability and decreased time, for making the measurements.

KW - 3D segmentation

KW - CT images

KW - level set

KW - Lung tumors

KW - sparse field active surface

UR - http://www.scopus.com/inward/record.url?scp=84856976617&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84856976617&partnerID=8YFLogxK

U2 - 10.1118/1.3676687

DO - 10.1118/1.3676687

M3 - Article

C2 - 22320795

AN - SCOPUS:84856976617

VL - 39

SP - 851

EP - 865

JO - Medical Physics

JF - Medical Physics

SN - 0094-2405

IS - 2

ER -