Automated individual-level parcellation of Broca's region based on functional connectivity

Estrid Jakobsen, Franziskus Liem, Manousos A Klados, Şeyma Bayrak, Michael Petrides, Daniel S Margulies

Research output: Contribution to journalArticle

Abstract

Broca's region can be subdivided into its constituent areas 44 and 45 based on established differences in connectivity to superior temporal and inferior parietal regions. The current study builds on our previous work manually parcellating Broca's area on the individual-level by applying these anatomical criteria to functional connectivity data. Here we present an automated observer-independent and anatomy-informed parcellation pipeline with comparable precision to the manual labels at the individual-level. The method first extracts individualized connectivity templates of areas 44 and 45 by assigning to each surface vertex within the ventrolateral frontal cortex the partial correlation value of its functional connectivity to group-level templates of areas 44 and 45, accounting for other template connectivity patterns. To account for cross-subject variability in connectivity, the partial correlation procedure is then repeated using individual-level network templates, including individual-level connectivity from areas 44 and 45. Each node is finally labeled as area 44, 45, or neither, using a winner-take-all approach. The method also incorporates prior knowledge of anatomical location by weighting the results using spatial probability maps. The resulting area labels show a high degree of spatial overlap with the gold-standard manual labels, and group-average area maps are consistent with cytoarchitectonic probability maps of areas 44 and 45. To facilitate reproducibility and to demonstrate that the method can be applied to resting-state fMRI datasets with varying acquisition and preprocessing parameters, the labeling procedure is applied to two open-source datasets from the Human Connectome Project and the Nathan Kline Institute Rockland Sample. While the current study focuses on Broca's region, the method is adaptable to parcellate other cortical regions with distinct connectivity profiles.

LanguageEnglish
Pages41-53
JournalNeuroImage
Volume170
Early online date30 Sep 2016
DOIs
Publication statusE-pub ahead of print - 30 Sep 2016

Fingerprint

Connectome
Parietal Lobe
Frontal Lobe
Gold
Broca Area
Anatomy
Magnetic Resonance Imaging
Datasets

Bibliographical note

© 2016 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY license (http://creativecommons.org/licenses/BY/4.0/).

Keywords

  • FMRI
  • Neuroimaging
  • Cortical
  • Parcellation
  • Language

Cite this

Jakobsen, Estrid ; Liem, Franziskus ; Klados, Manousos A ; Bayrak, Şeyma ; Petrides, Michael ; Margulies, Daniel S. / Automated individual-level parcellation of Broca's region based on functional connectivity. In: NeuroImage. 2016 ; Vol. 170. pp. 41-53.
@article{054ca2d4b29a4db48b0b11df530d6f0e,
title = "Automated individual-level parcellation of Broca's region based on functional connectivity",
abstract = "Broca's region can be subdivided into its constituent areas 44 and 45 based on established differences in connectivity to superior temporal and inferior parietal regions. The current study builds on our previous work manually parcellating Broca's area on the individual-level by applying these anatomical criteria to functional connectivity data. Here we present an automated observer-independent and anatomy-informed parcellation pipeline with comparable precision to the manual labels at the individual-level. The method first extracts individualized connectivity templates of areas 44 and 45 by assigning to each surface vertex within the ventrolateral frontal cortex the partial correlation value of its functional connectivity to group-level templates of areas 44 and 45, accounting for other template connectivity patterns. To account for cross-subject variability in connectivity, the partial correlation procedure is then repeated using individual-level network templates, including individual-level connectivity from areas 44 and 45. Each node is finally labeled as area 44, 45, or neither, using a winner-take-all approach. The method also incorporates prior knowledge of anatomical location by weighting the results using spatial probability maps. The resulting area labels show a high degree of spatial overlap with the gold-standard manual labels, and group-average area maps are consistent with cytoarchitectonic probability maps of areas 44 and 45. To facilitate reproducibility and to demonstrate that the method can be applied to resting-state fMRI datasets with varying acquisition and preprocessing parameters, the labeling procedure is applied to two open-source datasets from the Human Connectome Project and the Nathan Kline Institute Rockland Sample. While the current study focuses on Broca's region, the method is adaptable to parcellate other cortical regions with distinct connectivity profiles.",
keywords = "FMRI, Neuroimaging, Cortical, Parcellation, Language",
author = "Estrid Jakobsen and Franziskus Liem and Klados, {Manousos A} and Şeyma Bayrak and Michael Petrides and Margulies, {Daniel S}",
note = "{\circledC} 2016 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY license (http://creativecommons.org/licenses/BY/4.0/).",
year = "2016",
month = "9",
day = "30",
doi = "10.1016/j.neuroimage.2016.09.069",
language = "English",
volume = "170",
pages = "41--53",
journal = "Neuroimage",
issn = "1053-8119",
publisher = "Elsevier",

}

Automated individual-level parcellation of Broca's region based on functional connectivity. / Jakobsen, Estrid; Liem, Franziskus; Klados, Manousos A; Bayrak, Şeyma; Petrides, Michael; Margulies, Daniel S.

In: NeuroImage, Vol. 170, 30.09.2016, p. 41-53.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Automated individual-level parcellation of Broca's region based on functional connectivity

AU - Jakobsen, Estrid

AU - Liem, Franziskus

AU - Klados, Manousos A

AU - Bayrak, Şeyma

AU - Petrides, Michael

AU - Margulies, Daniel S

N1 - © 2016 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY license (http://creativecommons.org/licenses/BY/4.0/).

PY - 2016/9/30

Y1 - 2016/9/30

N2 - Broca's region can be subdivided into its constituent areas 44 and 45 based on established differences in connectivity to superior temporal and inferior parietal regions. The current study builds on our previous work manually parcellating Broca's area on the individual-level by applying these anatomical criteria to functional connectivity data. Here we present an automated observer-independent and anatomy-informed parcellation pipeline with comparable precision to the manual labels at the individual-level. The method first extracts individualized connectivity templates of areas 44 and 45 by assigning to each surface vertex within the ventrolateral frontal cortex the partial correlation value of its functional connectivity to group-level templates of areas 44 and 45, accounting for other template connectivity patterns. To account for cross-subject variability in connectivity, the partial correlation procedure is then repeated using individual-level network templates, including individual-level connectivity from areas 44 and 45. Each node is finally labeled as area 44, 45, or neither, using a winner-take-all approach. The method also incorporates prior knowledge of anatomical location by weighting the results using spatial probability maps. The resulting area labels show a high degree of spatial overlap with the gold-standard manual labels, and group-average area maps are consistent with cytoarchitectonic probability maps of areas 44 and 45. To facilitate reproducibility and to demonstrate that the method can be applied to resting-state fMRI datasets with varying acquisition and preprocessing parameters, the labeling procedure is applied to two open-source datasets from the Human Connectome Project and the Nathan Kline Institute Rockland Sample. While the current study focuses on Broca's region, the method is adaptable to parcellate other cortical regions with distinct connectivity profiles.

AB - Broca's region can be subdivided into its constituent areas 44 and 45 based on established differences in connectivity to superior temporal and inferior parietal regions. The current study builds on our previous work manually parcellating Broca's area on the individual-level by applying these anatomical criteria to functional connectivity data. Here we present an automated observer-independent and anatomy-informed parcellation pipeline with comparable precision to the manual labels at the individual-level. The method first extracts individualized connectivity templates of areas 44 and 45 by assigning to each surface vertex within the ventrolateral frontal cortex the partial correlation value of its functional connectivity to group-level templates of areas 44 and 45, accounting for other template connectivity patterns. To account for cross-subject variability in connectivity, the partial correlation procedure is then repeated using individual-level network templates, including individual-level connectivity from areas 44 and 45. Each node is finally labeled as area 44, 45, or neither, using a winner-take-all approach. The method also incorporates prior knowledge of anatomical location by weighting the results using spatial probability maps. The resulting area labels show a high degree of spatial overlap with the gold-standard manual labels, and group-average area maps are consistent with cytoarchitectonic probability maps of areas 44 and 45. To facilitate reproducibility and to demonstrate that the method can be applied to resting-state fMRI datasets with varying acquisition and preprocessing parameters, the labeling procedure is applied to two open-source datasets from the Human Connectome Project and the Nathan Kline Institute Rockland Sample. While the current study focuses on Broca's region, the method is adaptable to parcellate other cortical regions with distinct connectivity profiles.

KW - FMRI

KW - Neuroimaging

KW - Cortical

KW - Parcellation

KW - Language

UR - https://www.sciencedirect.com/science/article/pii/S1053811916305468?via%3Dihub

U2 - 10.1016/j.neuroimage.2016.09.069

DO - 10.1016/j.neuroimage.2016.09.069

M3 - Article

VL - 170

SP - 41

EP - 53

JO - Neuroimage

T2 - Neuroimage

JF - Neuroimage

SN - 1053-8119

ER -