Limits...
Consensus between pipelines in structural brain networks.

Parker CS, Deligianni F, Cardoso MJ, Daga P, Modat M, Dayan M, Clark CA, Ourselin S, Clayden JD - PLoS ONE (2014)

Bottom Line: Similarity of network connections was compared between pipelines employing different atlases by merging parcels to a common and equivalent node scale.We found a high agreement between the networks across a range of fiber density thresholds.In addition, we identified a robust core of highly connected regions coinciding with a peak in similarity across network density thresholds, and replicated these results with atlases at different node scales.

View Article: PubMed Central - PubMed

Affiliation: Centre for Medical Image Computing, University College London, London, United Kingdom; Imaging and Biophysics Unit, UCL Institute of Child Health, London, United Kingdom.

ABSTRACT
Structural brain networks may be reconstructed from diffusion MRI tractography data and have great potential to further our understanding of the topological organisation of brain structure in health and disease. Network reconstruction is complex and involves a series of processesing methods including anatomical parcellation, registration, fiber orientation estimation and whole-brain fiber tractography. Methodological choices at each stage can affect the anatomical accuracy and graph theoretical properties of the reconstructed networks, meaning applying different combinations in a network reconstruction pipeline may produce substantially different networks. Furthermore, the choice of which connections are considered important is unclear. In this study, we assessed the similarity between structural networks obtained using two independent state-of-the-art reconstruction pipelines. We aimed to quantify network similarity and identify the core connections emerging most robustly in both pipelines. Similarity of network connections was compared between pipelines employing different atlases by merging parcels to a common and equivalent node scale. We found a high agreement between the networks across a range of fiber density thresholds. In addition, we identified a robust core of highly connected regions coinciding with a peak in similarity across network density thresholds, and replicated these results with atlases at different node scales. The binary network properties of these core connections were similar between pipelines but showed some differences in atlases across node scales. This study demonstrates the utility of applying multiple structural network reconstrution pipelines to diffusion data in order to identify the most important connections for further study.

Show MeSH
Representative cortical parcellations of P1 and P2 at the native and common node scale.Temporal lobe regions in P1 native scale parcellations (P1-44, far left) were merged, resulting in a lower scale parcellation (P1-34, middle right). Selected regions across the entire cerebral cortex in P2 native scale parcellations (P2-68, middle left) were merged (P2-34, see Fig. 1). This resulted in a common and anatomically equivalent parcellation scale of 34 nodes for both P1 and P2 networks.
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4214749&req=5

pone-0111262-g003: Representative cortical parcellations of P1 and P2 at the native and common node scale.Temporal lobe regions in P1 native scale parcellations (P1-44, far left) were merged, resulting in a lower scale parcellation (P1-34, middle right). Selected regions across the entire cerebral cortex in P2 native scale parcellations (P2-68, middle left) were merged (P2-34, see Fig. 1). This resulted in a common and anatomically equivalent parcellation scale of 34 nodes for both P1 and P2 networks.

Mentions: The pipelines employed atlases with a different number of brain regions, preventing a direct connection-wise comparison between them. Therefore, parcels in the native atlases were merged to a common node scale (Fig. 2). The number of merges was the minimum required to give correspondence between the atlases and resulted in 34 brain regions (17 per hemisphere). Parcels in the Desikan-Killiany atlas (P2) were merged across the entire cortex based on anatomical correspondence to their equivalent Hammers atlas (P1) parcels. For example, the pars opercularis, pars orbitalis and pars triangularis parcels in the native Desikan-Killiany atlas corresponded to the inferior frontal gyrus parcel in the native Hammers atlas and therefore in both of the merged atlases. The Desikan-Killiany and Hammers atlases differed fundamentally in temporal lobe regions, meaning an equivalent merging of parcels could not be found. Therefore, the temporal lobe is itself considered as a single node in both merged parcellations (Fig. 2 and 3). The merging process did not result in identical parcellations. The remaining differences in common scale parcellations were due to alternative border criteria as well as alternative parcellation algorithms. Therefore, in addition to the native Desikan-Killiany and Hammers atlases, we also obtained the two merged 34 node scale versions of each atlas for each subject (Fig. 3).


Consensus between pipelines in structural brain networks.

Parker CS, Deligianni F, Cardoso MJ, Daga P, Modat M, Dayan M, Clark CA, Ourselin S, Clayden JD - PLoS ONE (2014)

Representative cortical parcellations of P1 and P2 at the native and common node scale.Temporal lobe regions in P1 native scale parcellations (P1-44, far left) were merged, resulting in a lower scale parcellation (P1-34, middle right). Selected regions across the entire cerebral cortex in P2 native scale parcellations (P2-68, middle left) were merged (P2-34, see Fig. 1). This resulted in a common and anatomically equivalent parcellation scale of 34 nodes for both P1 and P2 networks.
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4214749&req=5

pone-0111262-g003: Representative cortical parcellations of P1 and P2 at the native and common node scale.Temporal lobe regions in P1 native scale parcellations (P1-44, far left) were merged, resulting in a lower scale parcellation (P1-34, middle right). Selected regions across the entire cerebral cortex in P2 native scale parcellations (P2-68, middle left) were merged (P2-34, see Fig. 1). This resulted in a common and anatomically equivalent parcellation scale of 34 nodes for both P1 and P2 networks.
Mentions: The pipelines employed atlases with a different number of brain regions, preventing a direct connection-wise comparison between them. Therefore, parcels in the native atlases were merged to a common node scale (Fig. 2). The number of merges was the minimum required to give correspondence between the atlases and resulted in 34 brain regions (17 per hemisphere). Parcels in the Desikan-Killiany atlas (P2) were merged across the entire cortex based on anatomical correspondence to their equivalent Hammers atlas (P1) parcels. For example, the pars opercularis, pars orbitalis and pars triangularis parcels in the native Desikan-Killiany atlas corresponded to the inferior frontal gyrus parcel in the native Hammers atlas and therefore in both of the merged atlases. The Desikan-Killiany and Hammers atlases differed fundamentally in temporal lobe regions, meaning an equivalent merging of parcels could not be found. Therefore, the temporal lobe is itself considered as a single node in both merged parcellations (Fig. 2 and 3). The merging process did not result in identical parcellations. The remaining differences in common scale parcellations were due to alternative border criteria as well as alternative parcellation algorithms. Therefore, in addition to the native Desikan-Killiany and Hammers atlases, we also obtained the two merged 34 node scale versions of each atlas for each subject (Fig. 3).

Bottom Line: Similarity of network connections was compared between pipelines employing different atlases by merging parcels to a common and equivalent node scale.We found a high agreement between the networks across a range of fiber density thresholds.In addition, we identified a robust core of highly connected regions coinciding with a peak in similarity across network density thresholds, and replicated these results with atlases at different node scales.

View Article: PubMed Central - PubMed

Affiliation: Centre for Medical Image Computing, University College London, London, United Kingdom; Imaging and Biophysics Unit, UCL Institute of Child Health, London, United Kingdom.

ABSTRACT
Structural brain networks may be reconstructed from diffusion MRI tractography data and have great potential to further our understanding of the topological organisation of brain structure in health and disease. Network reconstruction is complex and involves a series of processesing methods including anatomical parcellation, registration, fiber orientation estimation and whole-brain fiber tractography. Methodological choices at each stage can affect the anatomical accuracy and graph theoretical properties of the reconstructed networks, meaning applying different combinations in a network reconstruction pipeline may produce substantially different networks. Furthermore, the choice of which connections are considered important is unclear. In this study, we assessed the similarity between structural networks obtained using two independent state-of-the-art reconstruction pipelines. We aimed to quantify network similarity and identify the core connections emerging most robustly in both pipelines. Similarity of network connections was compared between pipelines employing different atlases by merging parcels to a common and equivalent node scale. We found a high agreement between the networks across a range of fiber density thresholds. In addition, we identified a robust core of highly connected regions coinciding with a peak in similarity across network density thresholds, and replicated these results with atlases at different node scales. The binary network properties of these core connections were similar between pipelines but showed some differences in atlases across node scales. This study demonstrates the utility of applying multiple structural network reconstrution pipelines to diffusion data in order to identify the most important connections for further study.

Show MeSH