Automated PET attenuation correction model for functional brain imaging

B. T. Weinzapfel, Gary Hutchins

Research output: Contribution to journalArticle

27 Citations (Scopus)

Abstract

The failure to compensate for subject motion between attenuation correction scans and emission scans precludes the optimization of functional brain imaging techniques. We have developed an automated method for attenuation correction that compensates for subject motion by deriving each set of correction factors from the corresponding emission study. Methods: The technique consists of generation of an estimated skull image by filtered backprojection of the reciprocal of an emission sinogram; estimation of the thickness and radius of the skull on profiles extracted from the image; scaling the radius and thickness values to generate a model of the brain, skull, and scalp; and assignment of attenuation coefficients to the head model for generation of attenuation correction factors. Values for scale factors and tissue attenuation coefficients were determined empirically by fitting the emission-derived head model to measured transmission data in five subjects using nonlinear regression (group A). The average model parameters, across five datasets (group A), were then used to generate attenuation maps for five independent emission studies (group B). Mean-squared-error values were calculated between the measured transmission data and the two model groups. For comparison, mean squared error values were calculated between the measured transmission data and homogeneous ellipses that were manually fitted to emission images. Results: The difference between the mean squared error for groups A and B was not significant (P > 0.8), indicating that model parameters from a small group can be used for other subjects without further fitting. The mean squared error for the automated method was significantly lower than that of the ellipse method (P < 0.001). The method reduced emission image variance, resulting in a higher peak Z value in activation images. The elimination of measured transmission scans resulted in a reduction in scan time (∼15 min) and radiation exposure (∼0.5-1.6 mrem). Conclusion: We have developed an automated attenuation correction method that compensates for subject motion between scans, accurately reproduces the characteristics of the head, and eliminates the use of measured transmission data to reduce scan duration, statistical noise propagation, and radiation dose.

Original languageEnglish
Pages (from-to)483-491
Number of pages9
JournalJournal of Nuclear Medicine
Volume42
Issue number3
StatePublished - 2001

Fingerprint

Functional Neuroimaging
Skull
Head
Thromboplastin
Scalp
Radiation
Brain

Keywords

  • Attenuation correction
  • Brain mapping
  • CT, emission
  • Image processing, computer-assisted
  • PET

ASJC Scopus subject areas

  • Radiological and Ultrasound Technology

Cite this

Automated PET attenuation correction model for functional brain imaging. / Weinzapfel, B. T.; Hutchins, Gary.

In: Journal of Nuclear Medicine, Vol. 42, No. 3, 2001, p. 483-491.

Research output: Contribution to journalArticle

@article{64076f93f1f545b68fe3a18f64ebf7f2,
title = "Automated PET attenuation correction model for functional brain imaging",
abstract = "The failure to compensate for subject motion between attenuation correction scans and emission scans precludes the optimization of functional brain imaging techniques. We have developed an automated method for attenuation correction that compensates for subject motion by deriving each set of correction factors from the corresponding emission study. Methods: The technique consists of generation of an estimated skull image by filtered backprojection of the reciprocal of an emission sinogram; estimation of the thickness and radius of the skull on profiles extracted from the image; scaling the radius and thickness values to generate a model of the brain, skull, and scalp; and assignment of attenuation coefficients to the head model for generation of attenuation correction factors. Values for scale factors and tissue attenuation coefficients were determined empirically by fitting the emission-derived head model to measured transmission data in five subjects using nonlinear regression (group A). The average model parameters, across five datasets (group A), were then used to generate attenuation maps for five independent emission studies (group B). Mean-squared-error values were calculated between the measured transmission data and the two model groups. For comparison, mean squared error values were calculated between the measured transmission data and homogeneous ellipses that were manually fitted to emission images. Results: The difference between the mean squared error for groups A and B was not significant (P > 0.8), indicating that model parameters from a small group can be used for other subjects without further fitting. The mean squared error for the automated method was significantly lower than that of the ellipse method (P < 0.001). The method reduced emission image variance, resulting in a higher peak Z value in activation images. The elimination of measured transmission scans resulted in a reduction in scan time (∼15 min) and radiation exposure (∼0.5-1.6 mrem). Conclusion: We have developed an automated attenuation correction method that compensates for subject motion between scans, accurately reproduces the characteristics of the head, and eliminates the use of measured transmission data to reduce scan duration, statistical noise propagation, and radiation dose.",
keywords = "Attenuation correction, Brain mapping, CT, emission, Image processing, computer-assisted, PET",
author = "Weinzapfel, {B. T.} and Gary Hutchins",
year = "2001",
language = "English",
volume = "42",
pages = "483--491",
journal = "Journal of Nuclear Medicine",
issn = "0161-5505",
publisher = "Society of Nuclear Medicine Inc.",
number = "3",

}

TY - JOUR

T1 - Automated PET attenuation correction model for functional brain imaging

AU - Weinzapfel, B. T.

AU - Hutchins, Gary

PY - 2001

Y1 - 2001

N2 - The failure to compensate for subject motion between attenuation correction scans and emission scans precludes the optimization of functional brain imaging techniques. We have developed an automated method for attenuation correction that compensates for subject motion by deriving each set of correction factors from the corresponding emission study. Methods: The technique consists of generation of an estimated skull image by filtered backprojection of the reciprocal of an emission sinogram; estimation of the thickness and radius of the skull on profiles extracted from the image; scaling the radius and thickness values to generate a model of the brain, skull, and scalp; and assignment of attenuation coefficients to the head model for generation of attenuation correction factors. Values for scale factors and tissue attenuation coefficients were determined empirically by fitting the emission-derived head model to measured transmission data in five subjects using nonlinear regression (group A). The average model parameters, across five datasets (group A), were then used to generate attenuation maps for five independent emission studies (group B). Mean-squared-error values were calculated between the measured transmission data and the two model groups. For comparison, mean squared error values were calculated between the measured transmission data and homogeneous ellipses that were manually fitted to emission images. Results: The difference between the mean squared error for groups A and B was not significant (P > 0.8), indicating that model parameters from a small group can be used for other subjects without further fitting. The mean squared error for the automated method was significantly lower than that of the ellipse method (P < 0.001). The method reduced emission image variance, resulting in a higher peak Z value in activation images. The elimination of measured transmission scans resulted in a reduction in scan time (∼15 min) and radiation exposure (∼0.5-1.6 mrem). Conclusion: We have developed an automated attenuation correction method that compensates for subject motion between scans, accurately reproduces the characteristics of the head, and eliminates the use of measured transmission data to reduce scan duration, statistical noise propagation, and radiation dose.

AB - The failure to compensate for subject motion between attenuation correction scans and emission scans precludes the optimization of functional brain imaging techniques. We have developed an automated method for attenuation correction that compensates for subject motion by deriving each set of correction factors from the corresponding emission study. Methods: The technique consists of generation of an estimated skull image by filtered backprojection of the reciprocal of an emission sinogram; estimation of the thickness and radius of the skull on profiles extracted from the image; scaling the radius and thickness values to generate a model of the brain, skull, and scalp; and assignment of attenuation coefficients to the head model for generation of attenuation correction factors. Values for scale factors and tissue attenuation coefficients were determined empirically by fitting the emission-derived head model to measured transmission data in five subjects using nonlinear regression (group A). The average model parameters, across five datasets (group A), were then used to generate attenuation maps for five independent emission studies (group B). Mean-squared-error values were calculated between the measured transmission data and the two model groups. For comparison, mean squared error values were calculated between the measured transmission data and homogeneous ellipses that were manually fitted to emission images. Results: The difference between the mean squared error for groups A and B was not significant (P > 0.8), indicating that model parameters from a small group can be used for other subjects without further fitting. The mean squared error for the automated method was significantly lower than that of the ellipse method (P < 0.001). The method reduced emission image variance, resulting in a higher peak Z value in activation images. The elimination of measured transmission scans resulted in a reduction in scan time (∼15 min) and radiation exposure (∼0.5-1.6 mrem). Conclusion: We have developed an automated attenuation correction method that compensates for subject motion between scans, accurately reproduces the characteristics of the head, and eliminates the use of measured transmission data to reduce scan duration, statistical noise propagation, and radiation dose.

KW - Attenuation correction

KW - Brain mapping

KW - CT, emission

KW - Image processing, computer-assisted

KW - PET

UR - http://www.scopus.com/inward/record.url?scp=0035092988&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0035092988&partnerID=8YFLogxK

M3 - Article

C2 - 11337527

AN - SCOPUS:0035092988

VL - 42

SP - 483

EP - 491

JO - Journal of Nuclear Medicine

JF - Journal of Nuclear Medicine

SN - 0161-5505

IS - 3

ER -