Evaluation of GMI and PMI diffeomorphic-based demons algorithms for aligning PET and CT Images

Juan Yang, Hongjun Wang, You Zhang, Yong Yin

Research output: Contribution to journalArticle

Abstract

Fusion of anatomic information in computed tomography (CT) and functional information in 18 F-FDG positron emission tomography (PET) is crucial for accurate differentiation of tumor from benign masses, designing radiotherapy treatment plan and staging of cancer. Although current PET and CT images can be acquired from combined 18 F-FDG PET/CT scanner, the two acquisitions are scanned separately and take a long time, which may induce potential positional errors in global and local caused by respiratory motion or organ peristalsis. So registration (alignment) of whole-body PET and CT images is a prerequisite for their meaningful fusion. The purpose of this study was to assess the performance of two multimodal registration algorithms for aligning PET and CT images. The proposed gradient of mutual information (GMI)-based demons algorithm, which incorporated the GMI between two images as an external force to facilitate the alignment, was compared with the point-wise mutual information (PMI) diffeomorphic-based demons algorithm whose external force was modified by replacing the image intensity difference in diffeomorphic demons algorithm with the PMI to make it appropriate for multimodal image registration. Eight patients with esophageal cancer(s) were enrolled in this IRB-approved study. Whole-body PET and CT images were acquired from a combined 18F-FDG PET/CT scanner for each patient. The modified Hausdorff distance (d MH ) was used to evaluate the registration accuracy of the two algorithms. Of all patients, the mean values and standard deviations (SDs) of d MH were 6.65 (± 1.90) voxels and 6.01 (± 1.90) after the GMI-based demons and the PMI diffeomorphic-based demons registration algorithms respectively. Preliminary results on oncological patients showed that the respiratory motion and organ peristalsis in PET/CT esophageal images could not be neglected, although a combined 18 F-FDG PET/CT scanner was used for image acquisition. The PMI diffeomorphic-based demons algorithm was more accurate than the GMI-based demons algorithm in registering PET/CT esophageal images.

Original languageEnglish (US)
Pages (from-to)18-30
Number of pages13
JournalJournal of applied clinical medical physics
Volume16
Issue number4
DOIs
StatePublished - Jan 1 2015
Externally publishedYes

Fingerprint

Positron emission tomography
Tomography
positrons
tomography
gradients
evaluation
X-Ray Computed Tomography Scanners
Peristalsis
Fusion reactions
Positron Emission Tomography Computed Tomography
scanners
Neoplasm Staging
Research Ethics Committees
Fluorodeoxyglucose F18
Esophageal Neoplasms
Image acquisition
Image registration
Radiotherapy
Positron-Emission Tomography
organs

Keywords

  • Deformable image registration
  • Demons algorithm
  • Multimodal
  • Mutual information
  • PET/CT

ASJC Scopus subject areas

  • Radiation
  • Instrumentation
  • Radiology Nuclear Medicine and imaging

Cite this

Evaluation of GMI and PMI diffeomorphic-based demons algorithms for aligning PET and CT Images. / Yang, Juan; Wang, Hongjun; Zhang, You; Yin, Yong.

In: Journal of applied clinical medical physics, Vol. 16, No. 4, 01.01.2015, p. 18-30.

Research output: Contribution to journalArticle

@article{0546215cf34b4fb8a6d380949dacbe52,
title = "Evaluation of GMI and PMI diffeomorphic-based demons algorithms for aligning PET and CT Images",
abstract = "Fusion of anatomic information in computed tomography (CT) and functional information in 18 F-FDG positron emission tomography (PET) is crucial for accurate differentiation of tumor from benign masses, designing radiotherapy treatment plan and staging of cancer. Although current PET and CT images can be acquired from combined 18 F-FDG PET/CT scanner, the two acquisitions are scanned separately and take a long time, which may induce potential positional errors in global and local caused by respiratory motion or organ peristalsis. So registration (alignment) of whole-body PET and CT images is a prerequisite for their meaningful fusion. The purpose of this study was to assess the performance of two multimodal registration algorithms for aligning PET and CT images. The proposed gradient of mutual information (GMI)-based demons algorithm, which incorporated the GMI between two images as an external force to facilitate the alignment, was compared with the point-wise mutual information (PMI) diffeomorphic-based demons algorithm whose external force was modified by replacing the image intensity difference in diffeomorphic demons algorithm with the PMI to make it appropriate for multimodal image registration. Eight patients with esophageal cancer(s) were enrolled in this IRB-approved study. Whole-body PET and CT images were acquired from a combined 18F-FDG PET/CT scanner for each patient. The modified Hausdorff distance (d MH ) was used to evaluate the registration accuracy of the two algorithms. Of all patients, the mean values and standard deviations (SDs) of d MH were 6.65 (± 1.90) voxels and 6.01 (± 1.90) after the GMI-based demons and the PMI diffeomorphic-based demons registration algorithms respectively. Preliminary results on oncological patients showed that the respiratory motion and organ peristalsis in PET/CT esophageal images could not be neglected, although a combined 18 F-FDG PET/CT scanner was used for image acquisition. The PMI diffeomorphic-based demons algorithm was more accurate than the GMI-based demons algorithm in registering PET/CT esophageal images.",
keywords = "Deformable image registration, Demons algorithm, Multimodal, Mutual information, PET/CT",
author = "Juan Yang and Hongjun Wang and You Zhang and Yong Yin",
year = "2015",
month = "1",
day = "1",
doi = "10.1120/jacmp.v16i4.5148",
language = "English (US)",
volume = "16",
pages = "18--30",
journal = "Journal of Applied Clinical Medical Physics",
issn = "1526-9914",
publisher = "American Institute of Physics Publising LLC",
number = "4",

}

TY - JOUR

T1 - Evaluation of GMI and PMI diffeomorphic-based demons algorithms for aligning PET and CT Images

AU - Yang, Juan

AU - Wang, Hongjun

AU - Zhang, You

AU - Yin, Yong

PY - 2015/1/1

Y1 - 2015/1/1

N2 - Fusion of anatomic information in computed tomography (CT) and functional information in 18 F-FDG positron emission tomography (PET) is crucial for accurate differentiation of tumor from benign masses, designing radiotherapy treatment plan and staging of cancer. Although current PET and CT images can be acquired from combined 18 F-FDG PET/CT scanner, the two acquisitions are scanned separately and take a long time, which may induce potential positional errors in global and local caused by respiratory motion or organ peristalsis. So registration (alignment) of whole-body PET and CT images is a prerequisite for their meaningful fusion. The purpose of this study was to assess the performance of two multimodal registration algorithms for aligning PET and CT images. The proposed gradient of mutual information (GMI)-based demons algorithm, which incorporated the GMI between two images as an external force to facilitate the alignment, was compared with the point-wise mutual information (PMI) diffeomorphic-based demons algorithm whose external force was modified by replacing the image intensity difference in diffeomorphic demons algorithm with the PMI to make it appropriate for multimodal image registration. Eight patients with esophageal cancer(s) were enrolled in this IRB-approved study. Whole-body PET and CT images were acquired from a combined 18F-FDG PET/CT scanner for each patient. The modified Hausdorff distance (d MH ) was used to evaluate the registration accuracy of the two algorithms. Of all patients, the mean values and standard deviations (SDs) of d MH were 6.65 (± 1.90) voxels and 6.01 (± 1.90) after the GMI-based demons and the PMI diffeomorphic-based demons registration algorithms respectively. Preliminary results on oncological patients showed that the respiratory motion and organ peristalsis in PET/CT esophageal images could not be neglected, although a combined 18 F-FDG PET/CT scanner was used for image acquisition. The PMI diffeomorphic-based demons algorithm was more accurate than the GMI-based demons algorithm in registering PET/CT esophageal images.

AB - Fusion of anatomic information in computed tomography (CT) and functional information in 18 F-FDG positron emission tomography (PET) is crucial for accurate differentiation of tumor from benign masses, designing radiotherapy treatment plan and staging of cancer. Although current PET and CT images can be acquired from combined 18 F-FDG PET/CT scanner, the two acquisitions are scanned separately and take a long time, which may induce potential positional errors in global and local caused by respiratory motion or organ peristalsis. So registration (alignment) of whole-body PET and CT images is a prerequisite for their meaningful fusion. The purpose of this study was to assess the performance of two multimodal registration algorithms for aligning PET and CT images. The proposed gradient of mutual information (GMI)-based demons algorithm, which incorporated the GMI between two images as an external force to facilitate the alignment, was compared with the point-wise mutual information (PMI) diffeomorphic-based demons algorithm whose external force was modified by replacing the image intensity difference in diffeomorphic demons algorithm with the PMI to make it appropriate for multimodal image registration. Eight patients with esophageal cancer(s) were enrolled in this IRB-approved study. Whole-body PET and CT images were acquired from a combined 18F-FDG PET/CT scanner for each patient. The modified Hausdorff distance (d MH ) was used to evaluate the registration accuracy of the two algorithms. Of all patients, the mean values and standard deviations (SDs) of d MH were 6.65 (± 1.90) voxels and 6.01 (± 1.90) after the GMI-based demons and the PMI diffeomorphic-based demons registration algorithms respectively. Preliminary results on oncological patients showed that the respiratory motion and organ peristalsis in PET/CT esophageal images could not be neglected, although a combined 18 F-FDG PET/CT scanner was used for image acquisition. The PMI diffeomorphic-based demons algorithm was more accurate than the GMI-based demons algorithm in registering PET/CT esophageal images.

KW - Deformable image registration

KW - Demons algorithm

KW - Multimodal

KW - Mutual information

KW - PET/CT

UR - http://www.scopus.com/inward/record.url?scp=84938930303&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84938930303&partnerID=8YFLogxK

U2 - 10.1120/jacmp.v16i4.5148

DO - 10.1120/jacmp.v16i4.5148

M3 - Article

C2 - 26218993

AN - SCOPUS:84938930303

VL - 16

SP - 18

EP - 30

JO - Journal of Applied Clinical Medical Physics

JF - Journal of Applied Clinical Medical Physics

SN - 1526-9914

IS - 4

ER -