Accelerated spatially encoded spectroscopy of arbitrarily shaped compartments using prior knowledge and linear

Algebraic Modeling, Paul A Bottomley, Yi Zhang

Research output: Contribution to journalArticle

Abstract

Low signal-to-noise ratios (SNR) and long scan times are endemic in spatially localized spectroscopy. While chemical shift imaging (CSI) offers the best SNR efficiency, the minimum scan time is set by the need to acquire a full set of phase-encoding gradients steps (PEs), even if an average spectrum from a multi-voxel compartment would suffice. Moreover, adding voxels to create a compartment-average spectrum post-acquisition sacrifices SNR compared to directly acquiring the spectrum from a correctly sized compartment. Spectroscopy with linear algebraic modeling (SLAM) is a relatively new approach to localization that directly encodes spectra in C arbitrarily prescribed compartments that are segmented from accompanying scout images. SLAM uses a tiny subset (≥C) of the CSI PEs selected from central k-space plus prior knowledge of the compartments, to maximize SNR/volume and greatly reduce scan times. Spectra representing compartment averages are reconstructed by solving a set of C linear simultaneous equations that eliminate unneeded PEs from the standard CSI algorithm. SLAM is further optimized for SNR and compartmental bleed errors, by permitting fractional- instead of integer-stepped gradients. SLAM is amenable to spatial and temporal inhomogeneity corrections, as used routinely in CSI. Combining SLAM with sensitivity encoding (SENSE) techniques from parallel MRI permits replacement of even more of the already-reduced SENSE encoding PEs, for even faster acquisitions. Adoption of the technique in initial cardiac 31P and brain 1H patient studies has demonstrated effective acceleration factors of 4- to 120-fold, SNR efficiency gains of several-fold, and accurate metabolite quantification with respect to reference CSI scans.

Original languageEnglish (US)
Pages (from-to)89-104
Number of pages16
JournaleMagRes
Volume4
Issue number1
DOIs
StatePublished - 2015

Fingerprint

Signal-To-Noise Ratio
Spectrum Analysis
Signal to noise ratio
Spectroscopy
Magnetic Resonance Imaging
Metabolites
Magnetic resonance imaging
Brain

Keywords

  • chemical shift imaging
  • human studies
  • localization
  • pulse sequences
  • reconstruction techniques
  • SLAM
  • spectroscopy

ASJC Scopus subject areas

  • Analytical Chemistry
  • Spectroscopy
  • Biomedical Engineering
  • Biochemistry
  • Radiology Nuclear Medicine and imaging

Cite this

Accelerated spatially encoded spectroscopy of arbitrarily shaped compartments using prior knowledge and linear. / Modeling, Algebraic; Bottomley, Paul A; Zhang, Yi.

In: eMagRes, Vol. 4, No. 1, 2015, p. 89-104.

Research output: Contribution to journalArticle

@article{3c7871b5e9b54aea9a55711e68a34e1f,
title = "Accelerated spatially encoded spectroscopy of arbitrarily shaped compartments using prior knowledge and linear",
abstract = "Low signal-to-noise ratios (SNR) and long scan times are endemic in spatially localized spectroscopy. While chemical shift imaging (CSI) offers the best SNR efficiency, the minimum scan time is set by the need to acquire a full set of phase-encoding gradients steps (PEs), even if an average spectrum from a multi-voxel compartment would suffice. Moreover, adding voxels to create a compartment-average spectrum post-acquisition sacrifices SNR compared to directly acquiring the spectrum from a correctly sized compartment. Spectroscopy with linear algebraic modeling (SLAM) is a relatively new approach to localization that directly encodes spectra in C arbitrarily prescribed compartments that are segmented from accompanying scout images. SLAM uses a tiny subset (≥C) of the CSI PEs selected from central k-space plus prior knowledge of the compartments, to maximize SNR/volume and greatly reduce scan times. Spectra representing compartment averages are reconstructed by solving a set of C linear simultaneous equations that eliminate unneeded PEs from the standard CSI algorithm. SLAM is further optimized for SNR and compartmental bleed errors, by permitting fractional- instead of integer-stepped gradients. SLAM is amenable to spatial and temporal inhomogeneity corrections, as used routinely in CSI. Combining SLAM with sensitivity encoding (SENSE) techniques from parallel MRI permits replacement of even more of the already-reduced SENSE encoding PEs, for even faster acquisitions. Adoption of the technique in initial cardiac 31P and brain 1H patient studies has demonstrated effective acceleration factors of 4- to 120-fold, SNR efficiency gains of several-fold, and accurate metabolite quantification with respect to reference CSI scans.",
keywords = "chemical shift imaging, human studies, localization, pulse sequences, reconstruction techniques, SLAM, spectroscopy",
author = "Algebraic Modeling and Bottomley, {Paul A} and Yi Zhang",
year = "2015",
doi = "10.1002/9780470034590.emrstm1438",
language = "English (US)",
volume = "4",
pages = "89--104",
journal = "eMagRes",
issn = "2055-6101",
publisher = "Blackwell",
number = "1",

}

TY - JOUR

T1 - Accelerated spatially encoded spectroscopy of arbitrarily shaped compartments using prior knowledge and linear

AU - Modeling, Algebraic

AU - Bottomley, Paul A

AU - Zhang, Yi

PY - 2015

Y1 - 2015

N2 - Low signal-to-noise ratios (SNR) and long scan times are endemic in spatially localized spectroscopy. While chemical shift imaging (CSI) offers the best SNR efficiency, the minimum scan time is set by the need to acquire a full set of phase-encoding gradients steps (PEs), even if an average spectrum from a multi-voxel compartment would suffice. Moreover, adding voxels to create a compartment-average spectrum post-acquisition sacrifices SNR compared to directly acquiring the spectrum from a correctly sized compartment. Spectroscopy with linear algebraic modeling (SLAM) is a relatively new approach to localization that directly encodes spectra in C arbitrarily prescribed compartments that are segmented from accompanying scout images. SLAM uses a tiny subset (≥C) of the CSI PEs selected from central k-space plus prior knowledge of the compartments, to maximize SNR/volume and greatly reduce scan times. Spectra representing compartment averages are reconstructed by solving a set of C linear simultaneous equations that eliminate unneeded PEs from the standard CSI algorithm. SLAM is further optimized for SNR and compartmental bleed errors, by permitting fractional- instead of integer-stepped gradients. SLAM is amenable to spatial and temporal inhomogeneity corrections, as used routinely in CSI. Combining SLAM with sensitivity encoding (SENSE) techniques from parallel MRI permits replacement of even more of the already-reduced SENSE encoding PEs, for even faster acquisitions. Adoption of the technique in initial cardiac 31P and brain 1H patient studies has demonstrated effective acceleration factors of 4- to 120-fold, SNR efficiency gains of several-fold, and accurate metabolite quantification with respect to reference CSI scans.

AB - Low signal-to-noise ratios (SNR) and long scan times are endemic in spatially localized spectroscopy. While chemical shift imaging (CSI) offers the best SNR efficiency, the minimum scan time is set by the need to acquire a full set of phase-encoding gradients steps (PEs), even if an average spectrum from a multi-voxel compartment would suffice. Moreover, adding voxels to create a compartment-average spectrum post-acquisition sacrifices SNR compared to directly acquiring the spectrum from a correctly sized compartment. Spectroscopy with linear algebraic modeling (SLAM) is a relatively new approach to localization that directly encodes spectra in C arbitrarily prescribed compartments that are segmented from accompanying scout images. SLAM uses a tiny subset (≥C) of the CSI PEs selected from central k-space plus prior knowledge of the compartments, to maximize SNR/volume and greatly reduce scan times. Spectra representing compartment averages are reconstructed by solving a set of C linear simultaneous equations that eliminate unneeded PEs from the standard CSI algorithm. SLAM is further optimized for SNR and compartmental bleed errors, by permitting fractional- instead of integer-stepped gradients. SLAM is amenable to spatial and temporal inhomogeneity corrections, as used routinely in CSI. Combining SLAM with sensitivity encoding (SENSE) techniques from parallel MRI permits replacement of even more of the already-reduced SENSE encoding PEs, for even faster acquisitions. Adoption of the technique in initial cardiac 31P and brain 1H patient studies has demonstrated effective acceleration factors of 4- to 120-fold, SNR efficiency gains of several-fold, and accurate metabolite quantification with respect to reference CSI scans.

KW - chemical shift imaging

KW - human studies

KW - localization

KW - pulse sequences

KW - reconstruction techniques

KW - SLAM

KW - spectroscopy

UR - http://www.scopus.com/inward/record.url?scp=85019367147&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85019367147&partnerID=8YFLogxK

U2 - 10.1002/9780470034590.emrstm1438

DO - 10.1002/9780470034590.emrstm1438

M3 - Article

AN - SCOPUS:85019367147

VL - 4

SP - 89

EP - 104

JO - eMagRes

JF - eMagRes

SN - 2055-6101

IS - 1

ER -